# feature-store

A feature store is a centralized repository for storing, retrieving, managing, and sharing machine-learning features, while keeping feature computation consistent between model training and inference.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## How Medium Powers Real-Time Recommendations at 1M OPS

DevFeed: [How Medium Powers Real-Time Recommendations at 1M OPS](<https://devfeed.tech/articles/how-medium-powers-real-time-recommendations-at-1m-ops-129449.md>)

Original publisher: [Read original article](<https://hackernoon.com/how-medium-powers-real-time-recommendations-at-1m-ops?source=rss>)

Author: ScyllaDB

Published: 2026-11-11T12:17:12Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [database-optimization](<https://devfeed.tech/topics/database-optimization.md>), [scylladb](<https://devfeed.tech/topics/scylladb.md>), [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>)

Tags: [feature-store](<https://devfeed.tech/tags/feature-store.md>), [good-company](<https://devfeed.tech/tags/good-company.md>), [latency](<https://devfeed.tech/tags/latency.md>), [machine-learning-feature-store](<https://devfeed.tech/tags/machine-learning-feature-store.md>), [medium](<https://devfeed.tech/tags/medium.md>), [medium-scylladb-feature-store](<https://devfeed.tech/tags/medium-scylladb-feature-store.md>), [real-time-recommendations](<https://devfeed.tech/tags/real-time-recommendations.md>), [recommendations](<https://devfeed.tech/tags/recommendations.md>), [scylladb](<https://devfeed.tech/tags/scylladb.md>), [scylladb-data-modeling](<https://devfeed.tech/tags/scylladb-data-modeling.md>), [scylladb-feature-store](<https://devfeed.tech/tags/scylladb-feature-store.md>), [scylladb-list-features](<https://devfeed.tech/tags/scylladb-list-features.md>), [scylladb-ttl](<https://devfeed.tech/tags/scylladb-ttl.md>), [scylladb-vs-dynamodb-benchmark](<https://devfeed.tech/tags/scylladb-vs-dynamodb-benchmark.md>)

### AI overview

Medium replaced relational features split across tables with list features stored per entity in a single ScyllaDB partition. This turns reads of recent list items into one ordered, limited query instead of an ID lookup followed by a query per item; it also avoids ALLOW FILTERING, which had discarded 90% of rows read. Items sharing a timestamp use a value hash in the clustering key. Native per-row TTL expires old data automatically. Removing selected values still takes two steps: a local secondary index finds matching keys, then a batch deletes them. In benchmarks on production data, ScyllaDB had lower and steadier reported latencies than DynamoDB for adding and getting items, especially at P95 and P99. DynamoDB ran without an extra caching layer, and Medium said it lacked data to assess the cache's potential effect.

### Source excerpt

Discover how Medium redesigned its ScyllaDB feature store with list-based features to reduce queries, improve latency, & support up to 1M operations per second.

## From Data Silos to Feature Serving: How We Unified Fragmented Data for Training, Inference, and Reuse

DevFeed: [From Data Silos to Feature Serving: How We Unified Fragmented Data for Training, Inference, and Reuse](<https://devfeed.tech/articles/from-data-silos-to-feature-serving-how-we-unified-fragmented-data-for-training-inference-and-reuse-129390.md>)

Original publisher: [Read original article](<https://product.hubspot.com/blog/from-data-silos-to-feature-serving-how-we-unified-fragmented-data-for-training-inference-and-reuse>)

Author: Leena Bhandari

Published: 2026-11-10T20:17:40Z

Content type: article

Language: en

Sources: [HubSpot](<https://devfeed.tech/sources/hubspot.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [AI-Ready Data](<https://devfeed.tech/topics/ai-ready-data.md>)

Tags: [data](<https://devfeed.tech/tags/data.md>), [data-silos](<https://devfeed.tech/tags/data-silos.md>), [engineer](<https://devfeed.tech/tags/engineer.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [feature](<https://devfeed.tech/tags/feature.md>), [hubspot](<https://devfeed.tech/tags/hubspot.md>), [inference](<https://devfeed.tech/tags/inference.md>), [reuse](<https://devfeed.tech/tags/reuse.md>), [serving](<https://devfeed.tech/tags/serving.md>), [silos](<https://devfeed.tech/tags/silos.md>), [software](<https://devfeed.tech/tags/software.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

HubSpot's feature store unifies discovery and retrieval while letting teams keep data in its existing system when practical. Its registry maps feature definitions to sources; HBase serves ingested online values, CRM properties and team services can be queried directly, and Apache Iceberg supports offline history and training. Each online write emits an event that Spark stages on S3; a daily loader builds point-in-time Iceberg rows, correcting later rows when events arrive late and adding columns for new features automatically. Teams can also connect existing Iceberg tables without ingestion if they meet the required schema and append-history expectations and grant read-only access. Direct source serving avoids duplicate storage and sync pipelines, but external latency, availability, quality, coverage and lineage are harder to guarantee; offline query performance also depends on table layout. The article reports roughly 1.4 billion daily online feature ingestions and gives examples of model and product uses.

### Source excerpt

Written by Leena Bhandari, Senior Software Engineer II, Andrea Cicalese, Senior Software Engineer II, and Durai Ramasubramanian, Senior Software Engineer II at HubSpot.

## Real-time fraud detection with Feature Views

DevFeed: [Real-time fraud detection with Feature Views](<https://devfeed.tech/articles/real-time-fraud-detection-with-feature-views-121507.md>)

Original publisher: [Read original article](<https://www.databricks.com/blog/real-time-fraud-detection-feature-views>)

Author: Shyam Sankararaman; Hedvig Vahlberg

Published: 2026-11-05T23:23:36Z

Content type: article

Language: en

Sources: [Databricks](<https://devfeed.tech/sources/databricks.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [online-fraud-prevention](<https://devfeed.tech/topics/online-fraud-prevention.md>), [streaming data analytics](<https://devfeed.tech/topics/streaming-data-analytics.md>)

Tags: [aggregates](<https://devfeed.tech/tags/aggregates.md>), [aggregation](<https://devfeed.tech/tags/aggregation.md>), [api](<https://devfeed.tech/tags/api.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [batch](<https://devfeed.tech/tags/batch.md>), [customer](<https://devfeed.tech/tags/customer.md>), [dashboard](<https://devfeed.tech/tags/dashboard.md>), [data-leakage](<https://devfeed.tech/tags/data-leakage.md>), [data-science-and-ml](<https://devfeed.tech/tags/data-science-and-ml.md>), [databricks](<https://devfeed.tech/tags/databricks.md>), [databricks-apps](<https://devfeed.tech/tags/databricks-apps.md>), [demand](<https://devfeed.tech/tags/demand.md>), [feature](<https://devfeed.tech/tags/feature.md>), [features](<https://devfeed.tech/tags/features.md>), [fraud](<https://devfeed.tech/tags/fraud.md>), [fraud-detection](<https://devfeed.tech/tags/fraud-detection.md>), [platform](<https://devfeed.tech/tags/platform.md>), [real-time](<https://devfeed.tech/tags/real-time.md>)

### AI overview

Databricks Feature Views use shared feature definitions for point-in-time training data and online fraud scoring, reducing data leakage and training-serving skew. The walkthrough ingests Kafka transactions into an offline table, computes rolling and Sawtooth aggregates, and combines a live payment amount with a customer's historical average and standard deviation through a request-time feature. After evaluating features on historical data, teams can register and materialize them into the Online Feature Store for Model Serving. The text notes that online lookup uses the latest available customer features rather than a snapshot selected by the request timestamp, and recommends waiting two full days before production use of the Sawtooth baseline. Unity Catalog governs feature definitions and lineage; source and materialized tables still require SELECT permissions.

### Source excerpt

Real-time fraud detection depends on three kinds of features: what a customer did...

## Enterprise ML Architecture: Making Every Boundary Earn Its Place

DevFeed: [Enterprise ML Architecture: Making Every Boundary Earn Its Place](<https://devfeed.tech/articles/enterprise-ml-architecture-making-every-boundary-earn-its-place-119001.md>)

Original publisher: [Read original article](<https://medium.com/snowflake/enterprise-ml-architecture-making-every-boundary-earn-its-place-c0ffaac36fd4?source=rss----34b6daafc07---4>)

Author: Anupama Gangadhar

Published: 2026-11-04T18:21:34Z

Content type: article

Language: en

Sources: [Snowflake Builders](<https://devfeed.tech/sources/snowflake-builders.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Inference](<https://devfeed.tech/topics/inference.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [data-governance](<https://devfeed.tech/topics/data-governance.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [observability](<https://devfeed.tech/topics/observability.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [also](<https://devfeed.tech/tags/also.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [architectures](<https://devfeed.tech/tags/architectures.md>), [automatic](<https://devfeed.tech/tags/automatic.md>), [based](<https://devfeed.tech/tags/based.md>), [boundary](<https://devfeed.tech/tags/boundary.md>), [campaign](<https://devfeed.tech/tags/campaign.md>), [can](<https://devfeed.tech/tags/can.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [churn](<https://devfeed.tech/tags/churn.md>), [compute](<https://devfeed.tech/tags/compute.md>), [consider](<https://devfeed.tech/tags/consider.md>), [container](<https://devfeed.tech/tags/container.md>), [cost](<https://devfeed.tech/tags/cost.md>), [created](<https://devfeed.tech/tags/created.md>), [customer](<https://devfeed.tech/tags/customer.md>), [data-engineering](<https://devfeed.tech/tags/data-engineering.md>), [data-science](<https://devfeed.tech/tags/data-science.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [enterprise-architecture](<https://devfeed.tech/tags/enterprise-architecture.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [making](<https://devfeed.tech/tags/making.md>), [ml](<https://devfeed.tech/tags/ml.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>)

### AI overview

ML workload placement should follow measured boundary costs and operating needs, not a platform's feature list. The excerpt proposes seven tests: assess data movement; compare total cost, including operations and change; distinguish batch, online and specialised workloads; place reusable features; decide training, governance, execution and monitoring separately; follow where predictions are consumed; and check whether agentic workflows change the workload requirements. Snowflake may suit batch scoring near governed data, reusable features and Snowflake based consumption. Mature low latency endpoints or workloads needing specialised runtimes may remain elsewhere. Hybrid designs can combine Snowflake features and governance with external training or online serving, then return predictions and outcomes for observation. Choose to keep, integrate, consolidate or build new based on workload evidence such as volume, latency, utilisation, recovery, controls and cost; recommendations should remain conditional when evidence is incomplete.

### Source excerpt

Seven Workload Tests for Better Architecture Decisions Enterprise ML architectures carry history. Many were built when the data platform handled storage, transformation and analytics, while a separate ML platform supplied notebooks, training infrastructure, registries, deployment pipelines, online endpoints, monitoring and specialist compute. Those choices created skills, controls and production services that may have improved over years. Snowflake now provides capabilities across more stages of the ML lifecycle. These capabilities now include feature management, development and job execution, model governance, warehouse and container-based inference, observability, explainability and lineage. The platform can also work with models trained elsewhere and with inference performed outside Snowflake. These broader capabilities merit reassessment, not automatic migration. A platform gaining a capability is not, by itself, an architecture justification for moving a workload. Placement should follow workload evidence: data movement, execution pattern, latency, feature reuse, operating maturity, prediction consumption, governance and total cost. Here is a set of seven tests to structure that discussion and decision making. Test 1: Does data movement materially affect the workload? "The data is already in Snowflake" is not a sufficient reason to run a machine learning workload there. Locality matters when the volume, frequency and operational consequences of crossing the platform boundary are substantial. As an example, consider a churn model that scores 50 million customer records each night. Source data and features reside in Snowflake, and the probabilities return there for campaign segmentation. External scoring creates a round trip; extract a large feature population, transfer it to another service, score it, move the predictions back, reconcile the result and recover from failures across two platforms. Running inference closer to the governed data may remove meaningful

## Why Real-Time AI at Scale is So Hard

DevFeed: [Why Real-Time AI at Scale is So Hard](<https://devfeed.tech/articles/why-real-time-ai-at-scale-is-so-hard-106767.md>)

Original publisher: [Read original article](<https://www.scylladb.com/2026/10/29/why-real-time-ai-at-scale-is-so-hard/>)

Author: Cynthia Dunlop

Published: 2026-10-29T12:23:15Z

Content type: article

Language: en

Sources: [ScyllaDB](<https://devfeed.tech/sources/scylladb.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [AI Architecture](<https://devfeed.tech/topics/ai-architecture.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [Real-time AI interactions](<https://devfeed.tech/topics/real-time-ai-interactions.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-ml](<https://devfeed.tech/tags/ai-ml.md>), [ai-performance](<https://devfeed.tech/tags/ai-performance.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [community](<https://devfeed.tech/tags/community.md>), [concurrent](<https://devfeed.tech/tags/concurrent.md>), [data](<https://devfeed.tech/tags/data.md>), [data-pipeline](<https://devfeed.tech/tags/data-pipeline.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [inference](<https://devfeed.tech/tags/inference.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [latency](<https://devfeed.tech/tags/latency.md>), [load](<https://devfeed.tech/tags/load.md>), [lock](<https://devfeed.tech/tags/lock.md>), [ml](<https://devfeed.tech/tags/ml.md>), [model](<https://devfeed.tech/tags/model.md>), [operations](<https://devfeed.tech/tags/operations.md>), [patterns](<https://devfeed.tech/tags/patterns.md>), [performance](<https://devfeed.tech/tags/performance.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [production](<https://devfeed.tech/tags/production.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [scale](<https://devfeed.tech/tags/scale.md>), [vector-search](<https://devfeed.tech/tags/vector-search.md>)

### AI overview

At roughly 740,000 operations per second, feature reads queued behind concurrent writes in Tim Koopmans's trading app, pushing P99 latency to three seconds; the model was not the cause. The excerpt also links accuracy loss to stale features and degrading HNSW indexes, and describes contention when training and serving share hardware. It recommends monitoring freshness, backlog, and index recall; load testing beyond steady state; separating workloads; and planning retraining with blue-green deployments, canaries, and application-layer cutovers. As one vendor example, ScyllaDB describes isolating vector indexing as an asynchronous service and prioritizing latency-sensitive workloads. Its cited benchmark reports P99 under 10 ms at concurrency 300, while noting higher recall trades off against latency and throughput. These are lessons and claims from the supplied excerpt, not universal performance guarantees.

### Source excerpt

``` html Why Real-Time AI at Scale is So Hard The failure patterns that surface as your app ramps up, and how to avoid them Real-time AI at scale is harder than it looks. Pipelines that hum along in development routinely hit problems in production. It's always easy to blame the model for all your problems. But issues like rising latency and degrading accuracy can usually be traced back to the data pipeline. My colleague Tim Koopmans and I recently discussed what typically goes wrong with real-time AI at scale. After Tim shared some hard-fought lessons learned, we talked about how to avoid falling into these traps yourself - including the practices and infrastructure choices that can help you avoid them. You can watch the full video or read the key points below. Why AI performance fails at scale Tim learned the following real-time AI performance lessons the hard way: through fits of frustration while building an ML-based financial trading app. You can't dig yourself out of tail latency All too often, latency looks fine in testing, then a P99 spike surfaces under real concurrent load. For example, as Tim's app approached ~740K operations per second, its P99 latency skyrocketed to 3 seconds. "I kept blaming the model for being slow, but it turns out the model was fine," Tim explained. "It was just that the feature lookups were killing me." Each inference call was doing just a handful of reads, but those reads queued up [behind writes] under load. The average latencies seemed fine, but that P99 tail latency was just unacceptable. Once you hit highly concurrent write throughput, you get lock contention - and that impacts the tail latencies. At this point, retries and bigger caches and connection pool tuning don't help. As Tim put it, "Tail latency isn't a bug that you can fix, it's a property of your architecture. For example, if your storage engine is producing GC pauses at exactly the wrong moment, you're going to cop a latency spike, no matter what." The culprit in Ti

## Airbnb Cuts Recommendation Staleness from 2 Days to Seconds with Chronon

DevFeed: [Airbnb Cuts Recommendation Staleness from 2 Days to Seconds with Chronon](<https://devfeed.tech/articles/airbnb-cuts-recommendation-staleness-from-2-days-to-seconds-with-chronon-104516.md>)

Original publisher: [Read original article](<https://www.infoq.com/news/2026/10/airbnb-chronon-recommendation/>)

Author: Leela Kumili

Published: 2026-10-28T13:53:00Z

Content type: news

Language: en

Sources: [InfoQ](<https://devfeed.tech/sources/infoq.md>)

Topics: [streaming data analytics](<https://devfeed.tech/topics/streaming-data-analytics.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [personalization](<https://devfeed.tech/topics/personalization.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [a-b-testing](<https://devfeed.tech/tags/a-b-testing.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-ml-data-engineering](<https://devfeed.tech/tags/ai-ml-data-engineering.md>), [airbnb](<https://devfeed.tech/tags/airbnb.md>), [airbnb-chronon-recommendation](<https://devfeed.tech/tags/airbnb-chronon-recommendation.md>), [allows](<https://devfeed.tech/tags/allows.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [architecture-design](<https://devfeed.tech/tags/architecture-design.md>), [based](<https://devfeed.tech/tags/based.md>), [batch-processing](<https://devfeed.tech/tags/batch-processing.md>), [days](<https://devfeed.tech/tags/days.md>), [development](<https://devfeed.tech/tags/development.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [event-driven-architecture](<https://devfeed.tech/tags/event-driven-architecture.md>), [feature](<https://devfeed.tech/tags/feature.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml-data-engineering](<https://devfeed.tech/tags/ml-data-engineering.md>), [mlops](<https://devfeed.tech/tags/mlops.md>), [model](<https://devfeed.tech/tags/model.md>), [news](<https://devfeed.tech/tags/news.md>), [offline](<https://devfeed.tech/tags/offline.md>), [online](<https://devfeed.tech/tags/online.md>), [platform](<https://devfeed.tech/tags/platform.md>), [platforms](<https://devfeed.tech/tags/platforms.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [recommendation](<https://devfeed.tech/tags/recommendation.md>), [sequence](<https://devfeed.tech/tags/sequence.md>), [streaming](<https://devfeed.tech/tags/streaming.md>), [testing](<https://devfeed.tech/tags/testing.md>), [time](<https://devfeed.tech/tags/time.md>), [transformer-models](<https://devfeed.tech/tags/transformer-models.md>), [update](<https://devfeed.tech/tags/update.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

Airbnb added Push Mode and Near Real-Time Model Transform to Chronon so guest activity can trigger streaming inference with its existing JourneyFormer sequence encoder. The system combines updated events with guest history, writes refreshed embeddings to the existing low-latency store, and leaves the ranking path largely unchanged. Typical updates arrive in 10-30 seconds, reducing embedding staleness from nearly two days to under a minute. Airbnb reports a 1.67% offline NDCG improvement and about a 0.33% increase in uncancelled bookings in online A/B testing. The excerpt says this approach trades additional writes and pipeline complexity for fresher predictions; it contrasts Chronon's integrated streaming and model transformation with Feast's documented cached-prediction architecture. The features were contributed to Chronon's public repository.

### Source excerpt

Airbnb extended its Chronon feature platform with Push Mode and Near Real-Time Model Transform to update sequence-based recommendation embeddings within seconds. The architecture reduced embedding staleness from nearly two days to under a minute and improved offline NDCG by 1.67%, while online testing increased uncancelled bookings by approximately one-third of a percent. By Leela Kumili

## Feast Feature and Vector Serving at ScyllaDB Speed

DevFeed: [Feast Feature and Vector Serving at ScyllaDB Speed](<https://devfeed.tech/articles/feast-feature-and-vector-serving-at-scylladb-speed-89975.md>)

Original publisher: [Read original article](<https://www.scylladb.com/2026/10/19/feast-feature-and-vector-serving-at-scylladb-speed/>)

Author: Cynthia Dunlop

Published: 2026-10-19T12:12:42Z

Content type: tutorial

Language: en

Sources: [ScyllaDB](<https://devfeed.tech/sources/scylladb.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [scylladb](<https://devfeed.tech/topics/scylladb.md>), [Vectors & LLMs](<https://devfeed.tech/topics/vectors-llms.md>), [Latency](<https://devfeed.tech/topics/latency.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [Alternator (for DynamoDB)](<https://devfeed.tech/topics/alternator-for-dynamodb.md>), [Qdrant](<https://devfeed.tech/topics/qdrant.md>), [self-hosted databases vs managed services](<https://devfeed.tech/topics/self-hosted-databases-vs-managed-services.md>)

Tags: [ai-ml](<https://devfeed.tech/tags/ai-ml.md>), [api](<https://devfeed.tech/tags/api.md>), [availability](<https://devfeed.tech/tags/availability.md>), [cost](<https://devfeed.tech/tags/cost.md>), [data](<https://devfeed.tech/tags/data.md>), [database](<https://devfeed.tech/tags/database.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [distributed-architecture](<https://devfeed.tech/tags/distributed-architecture.md>), [ecosystem](<https://devfeed.tech/tags/ecosystem.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [performance](<https://devfeed.tech/tags/performance.md>), [scylladb](<https://devfeed.tech/tags/scylladb.md>), [vector-search](<https://devfeed.tech/tags/vector-search.md>)

### AI overview

The article describes the Feast v0.65 integration with ScyllaDB as an online feature store, adding a shard-aware Python driver, vector search support, and configuration through Feast. It explains feature-store scaling and operational considerations, then provides setup examples for feature retrieval and vector search.

### Source excerpt

A new and tighter integration between Feast and ScyllaDB online feature store, including vector search support and better performance

## Integrating Netflix's Foundation Model into Personalization applications

DevFeed: [Integrating Netflix's Foundation Model into Personalization applications](<https://devfeed.tech/articles/integrating-netflix-s-foundation-model-into-personalization-applications-87083.md>)

Original publisher: [Read original article](<https://netflixtechblog.com/integrating-netflixs-foundation-model-into-personalization-applications-cf176b5860eb?source=rss----2615bd06b42e---4>)

Author: Netflix Technology Blog

Published: 2026-10-14T20:02:22Z

Content type: tutorial

Language: en

Sources: [Netflix](<https://devfeed.tech/sources/netflix-techblog-medium.md>)

Topics: [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [personalization](<https://devfeed.tech/topics/personalization.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>), [AI for interactive applications](<https://devfeed.tech/topics/ai-for-interactive-applications.md>)

Tags: [applications](<https://devfeed.tech/tags/applications.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [fine-tuning](<https://devfeed.tech/tags/fine-tuning.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [integration](<https://devfeed.tech/tags/integration.md>), [latency](<https://devfeed.tech/tags/latency.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [netflix](<https://devfeed.tech/tags/netflix.md>), [personalization](<https://devfeed.tech/tags/personalization.md>), [production](<https://devfeed.tech/tags/production.md>), [transformer-model](<https://devfeed.tech/tags/transformer-model.md>)

### AI overview

Netflix describes three ways to integrate its personalization Foundation Model, chosen according to application latency needs, technology constraints, and desired depth of use. The simplest route is daily refreshed user and item embeddings in the Embedding Store: inexpensive and easy to add to existing models, but potentially stale. Embedding stabilization maps vectors across monthly retraining and daily fine-tuning; a near-real-time framework is being developed, though the text says it cannot serve very large models. Subgraph integration runs the Foundation Model decoder on raw interaction sequences as part of the application model, enabling fine-tuning and fresh representations, at the cost of more feature-generation work, compute, model size, and inference time. Netflix recommends reserving it for high-impact cases where metric gains justify that burden. Alternatively, teams can fine-tune the Foundation Model for their data and objective and use it directly, offering a baseline for new applications while adding models and pipelines to maintain and requiring latency and SLA optimization. The company is also developing a smaller distilled model to reduce subgraph latency.

### Source excerpt

Authors: Divya Gadde, Ko-Jen Hsiao, Dhaval Patel and Moumita Bhattacharya The motivation for building the Netflix Foundation Model for personalization was laid out in this previous blog post. The Netflix homepage is powered by several specialized models and it requires significant time and resources to maintain and continue innovating on these individual models. Therefore, we decided to centralize member preference learning and have one powerful model learned from comprehensive user interaction histories and content data at a large scale and have this model distribute its learnings to other models. The previous two blog posts(1, 2) focus on the development of the Foundation Model and here we talk about how we integrate the Foundation Model into various applications. We found that there is extensive literature focused on the training and inference of large-scale transformer-based models. However, there remains a significant gap in practical guidance and research on how to effectively and efficiently integrate large models into existing production systems. At Netflix, we experimented with 3 distinct integration approaches -- embeddings, subgraph and fine-tuning. The motivation for trying different approaches stems from varying application needs. Specifically, different teams have different latency requirements, tech stack constraints and different appetites for leveraging the full power of a large foundation model. In this blog, we go over the 3 approaches we experimented with and which are now used in production for different use cases. Each integration approach has its own set of motivations, tradeoffs, complexities and learnings. The following sections go into the details of each of these integration approaches. Embeddings from Foundation model Transformer style foundation models naturally lend themselves to generating comprehensive user profile and item representations. Our approach for generating embeddings includes getting the hidden state of the last user event

## Voting on Features: Federated Feature Selection Without Sharing Data

DevFeed: [Voting on Features: Federated Feature Selection Without Sharing Data](<https://devfeed.tech/articles/voting-on-features-federated-feature-selection-without-sharing-data-86947.md>)

Original publisher: [Read original article](<https://towardsdatascience.com/voting-on-features-federated-feature-selection-without-sharing-data/>)

Author: Ioannis Christofilogiannis

Published: 2026-10-14T12:30:00Z

Content type: tutorial

Language: en

Sources: [Towards Data Science](<https://devfeed.tech/sources/towards-data-science.md>)

Topics: [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [Federated Learning](<https://devfeed.tech/topics/federated-learning.md>), [NVIDIA Flare](<https://devfeed.tech/topics/nvidia-flare.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>)

Tags: [data](<https://devfeed.tech/tags/data.md>), [editor-s-picks](<https://devfeed.tech/tags/editor-s-picks.md>), [feature-engineering](<https://devfeed.tech/tags/feature-engineering.md>), [feature-selection](<https://devfeed.tech/tags/feature-selection.md>), [federated](<https://devfeed.tech/tags/federated.md>), [federated-learning](<https://devfeed.tech/tags/federated-learning.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [nvidia-flare](<https://devfeed.tech/tags/nvidia-flare.md>), [python](<https://devfeed.tech/tags/python.md>)

### AI overview

The article presents Feature Election, a method for federated feature selection in which sites keep their data local, run their own feature selectors, and send feature masks, scores, and sample counts to a server. The server combines weighted votes to choose contested features, then sites train on the resulting reduced feature set. It explains the method, demonstrates implementations in NVIDIA FLARE and Flower, and reviews reported experiments and an independent benchmark. The reported benefit is smaller models and less communication in many evaluations, while predictive results vary by dataset and model; masks and scores also provide no formal privacy guarantee.

### Source excerpt

Hospitals, banks, and edge devices can agree on which features matter by sharing votes instead of rows, runnable in NVIDIA FLARE and Flower The post Voting on Features: Federated Feature Selection Without Sharing Data appeared first on Towards Data Science.

## SN50 Premium Inference: A Faster ROI for Neoclouds

DevFeed: [SN50 Premium Inference: A Faster ROI for Neoclouds](<https://devfeed.tech/articles/sn50-premium-inference-a-faster-roi-for-neoclouds-81511.md>)

Original publisher: [Read original article](<https://sambanova.ai/blog/sn50-premium-inference-a-faster-roi-for-neoclouds>)

Author: SambaNova

Published: 2026-09-17T19:30:41Z

Content type: article

Language: en

Sources: [Sambanova](<https://devfeed.tech/sources/sambanova.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>)

Tags: [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [blog](<https://devfeed.tech/tags/blog.md>), [business](<https://devfeed.tech/tags/business.md>), [compute](<https://devfeed.tech/tags/compute.md>), [cost](<https://devfeed.tech/tags/cost.md>), [costs](<https://devfeed.tech/tags/costs.md>), [inference](<https://devfeed.tech/tags/inference.md>), [neoclouds](<https://devfeed.tech/tags/neoclouds.md>), [roi](<https://devfeed.tech/tags/roi.md>)

### AI overview

The article presents SambaNova's SN50 as a way for neocloud providers to offer faster inference for large models and coding-agent workloads. It pairs purpose-built RDU decode with compatible GPUs for prefill, and describes a six-month return-on-investment target based on recovering infrastructure costs from revenue after operating costs, using 70% billable utilization as an illustrative planning assumption.

### Source excerpt

AI agents are creating a premium inference business for neoclouds. Coding agents read a repository, generate changes, run tests, and return to the model with the results. As code, tool outputs, and prior turns accumulate, inputs can exceed 100,000 tokens. Artificial Analysis' AA-AgentPerf benchmark replays real coding-agent trajectories with inputs reaching approximately 131,000 tokens per request. Customers need large models that stay responsive as that work grows.

## Creating an AI Platform for classic ML online inference

DevFeed: [Creating an AI Platform for classic ML online inference](<https://devfeed.tech/articles/creating-an-ai-platform-for-classic-ml-online-inference-22589.md>)

Original publisher: [Read original article](<https://medium.com/amex-gbt-technology/creating-an-ai-platform-for-classic-ml-online-inference-e2165d68e18a?source=rss----60a0578f4096---4>)

Author: Rohith Leeladharan

Published: 2026-09-10T07:26:46Z

Content type: article

Language: en

Sources: [Amex GBT Technology](<https://devfeed.tech/sources/amex-gbt-technology.md>)

Topics: [Inference](<https://devfeed.tech/topics/inference.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [internal developer platform automation](<https://devfeed.tech/topics/internal-developer-platform-automation.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Advanced AI workflows](<https://devfeed.tech/topics/advanced-ai-workflows.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [Learning](<https://devfeed.tech/topics/learning.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-platform](<https://devfeed.tech/tags/ai-platform.md>), [ai-platform-engineering](<https://devfeed.tech/tags/ai-platform-engineering.md>), [continuous-improvement](<https://devfeed.tech/tags/continuous-improvement.md>), [data](<https://devfeed.tech/tags/data.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [events](<https://devfeed.tech/tags/events.md>), [experimentation](<https://devfeed.tech/tags/experimentation.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [inference](<https://devfeed.tech/tags/inference.md>), [inference-engine](<https://devfeed.tech/tags/inference-engine.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-models](<https://devfeed.tech/tags/ml-models.md>), [mlops](<https://devfeed.tech/tags/mlops.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [monitoring](<https://devfeed.tech/tags/monitoring.md>), [online](<https://devfeed.tech/tags/online.md>), [platform](<https://devfeed.tech/tags/platform.md>), [production](<https://devfeed.tech/tags/production.md>)

### AI overview

American Express Global Business Travel describes an online inference platform built around a pre-process, predict, post-process pattern: client requests go through OIS, which routes them to inference engines that can retrieve stored features, prepare inputs, run a model, and format outputs. An inference template and FastAPI-based InferenceApp handle production boilerplate such as authentication, contextual logs, and model downloads. OIS supports A/B, shadow, and user-subset testing, and asynchronously records requests and responses. To let model teams experiment without repeated client API changes, engines may accept anticipated future features; a benchmark with 20 unused features found under 1 ms of response-time overhead. The article reports 21 engines across 11 projects as of September 2026, with the Flight Ranking system handling millions of weekly requests and a 0.01% error rate. It also describes early latency problems and lists planned work, including event-driven inference, stronger drift monitoring, composable pipelines, and canary testing.

### Source excerpt

Introduction In 2021, we were given the mission to have AI Systems running in production. The team, instead of just following a classical MLOps process, that involves transforming a Jupyter notebook into a product running in production, decided to go further by creating a platform to deploy AI systems in production. The team decided the platform should respect these requirements: Simplicity: The code powering AI systems should be simple, readable, and easy to maintain -- less intricacy means fewer bugs in production and greater reliability. Self-service: Anyone should be able to build and deploy AI systems autonomously, without depending on a central team. Experimentation: The platform should make it easy to run and iterate on experiments. Continuous improvement: Data related to events and interactions within AI systems must be captured, enabling monitoring and continuous improvement over time. In this article, we will walk through the work done to build a platform that fulfills these four requirements. Background At American Express Global Business Travel, we use machine learning (ML) models for a variety of user experiences like ranking hotel and flight search results. Our ML models are wrapped in inference engines that handle both pre-processing of input data before we run a prediction with the model, and post-processing of output data before returning the output to the caller. The overall flow looks something like this: Figure 1: Handling an inference request A client service that would like the ML model's predictions provides necessary context about the request like which user the request is for. Then, optionally, the inference engine fetches any necessary features for inference from our feature store [part 1][part 2]. Finally, it pre-processes the data, runs the predictions using the trained ML model, and does any necessary post-processing of the model output before returning the response to the caller. We call this the pre-process, predict, post-process patter

## Audience Hub: Marketer-friendly features for audience activation | Blog | Fivetran

DevFeed: [Audience Hub: Marketer-friendly features for audience activation | Blog | Fivetran](<https://devfeed.tech/articles/audience-hub-marketer-friendly-features-for-audience-activation-blog-fivetran-78391.md>)

Original publisher: [Read original article](<https://www.fivetran.com/blog/audience-hub-marketer-friendly-features-for-audience-activation>)

Author: Sean Lynch

Published: 2026-07-22T17:00:57Z

Content type: article

Language: en

Sources: [Fivetran Blog](<https://devfeed.tech/sources/fivetran-blog.md>)

Topics: [Data Mining & Modeling](<https://devfeed.tech/topics/data-mining-modeling.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>)

Tags: [business](<https://devfeed.tech/tags/business.md>), [customer](<https://devfeed.tech/tags/customer.md>), [customer-360](<https://devfeed.tech/tags/customer-360.md>), [customer-data](<https://devfeed.tech/tags/customer-data.md>), [engagement](<https://devfeed.tech/tags/engagement.md>), [features](<https://devfeed.tech/tags/features.md>)

### AI overview

Fivetran Audience Hub is a warehouse-native tool for building and activating customer audiences without SQL. It offers visual segmentation, audience comparisons, analytics, and synchronization to marketing platforms while allowing data teams to retain governance over warehouse-resident data.

### Source excerpt

Empower every marketer to build, analyze, and activate precise audiences with a no-code, warehouse-native segmentation suite.

## Using Dolt's Versioned SQL Database for Feature-Store Point-in-Time Reads

DevFeed: [Using Dolt's Versioned SQL Database for Feature-Store Point-in-Time Reads](<https://devfeed.tech/articles/what-if-the-feature-store-had-git-built-in-40139.md>)

Original publisher: [Read original article](<https://korbonits.com/blog/2026-05-15-what-if-the-feature-store-had-git-built-in/>)

Published: 2026-05-15T00:00:00Z

Content type: opinion

Language: en

Sources: [Alex Korbonits](<https://devfeed.tech/sources/alex-korbonits.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [Git](<https://devfeed.tech/topics/git.md>), [reproducibility](<https://devfeed.tech/topics/reproducibility.md>), [Database](<https://devfeed.tech/topics/database.md>), [SQL](<https://devfeed.tech/topics/sql.md>)

Tags: [dedupe](<https://devfeed.tech/tags/dedupe.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [git](<https://devfeed.tech/tags/git.md>), [reproducibility](<https://devfeed.tech/tags/reproducibility.md>), [rfc](<https://devfeed.tech/tags/rfc.md>), [row-number](<https://devfeed.tech/tags/row-number.md>), [sql](<https://devfeed.tech/tags/sql.md>), [sql-database](<https://devfeed.tech/tags/sql-database.md>)

### AI overview

This article describes a prototype Feast offline-store plugin backed by Dolt, a version-controlled SQL database. It argues that Dolt's revision-based reads can provide point-in-time feature retrieval and reproducible training snapshots without the usual append-only log deduplication CTEs. Tests on toy datasets produced identical results with shorter queries, though the article presents this as an early spike rather than a production benchmark.

### Source excerpt

A weekend spike asked whether Dolt's AS OF reads could replace the ROW_NUMBER dedupe at the heart of every feature store's point-in-time join. Four weeks later, the RFC is quiet and the plugin's get_historical_features works end-to-end against a live Dolt server. Notes on building anyway.

## Engineering Features at Scale: Inside the Darwin Feature Store

DevFeed: [Engineering Features at Scale: Inside the Darwin Feature Store](<https://devfeed.tech/articles/engineering-features-at-scale-inside-the-darwin-feature-store-22619.md>)

Original publisher: [Read original article](<https://medium.com/dreamlockerroom/engineering-features-at-scale-inside-the-darwin-feature-store-ed5928752e8a?source=rss----5c7a7f580b01---4>)

Author: Dream Blog

Published: 2026-01-27T15:38:25Z

Content type: article

Language: en

Sources: [Dream11 Engineering](<https://devfeed.tech/sources/dream11-engineering.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [Latency](<https://devfeed.tech/topics/latency.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>)

Tags: [batch](<https://devfeed.tech/tags/batch.md>), [cassandra](<https://devfeed.tech/tags/cassandra.md>), [dream11](<https://devfeed.tech/tags/dream11.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [feature-engineering](<https://devfeed.tech/tags/feature-engineering.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [features](<https://devfeed.tech/tags/features.md>), [inference](<https://devfeed.tech/tags/inference.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [latency](<https://devfeed.tech/tags/latency.md>), [low-latency](<https://devfeed.tech/tags/low-latency.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-models](<https://devfeed.tech/tags/ml-models.md>), [offline](<https://devfeed.tech/tags/offline.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [platform](<https://devfeed.tech/tags/platform.md>), [production](<https://devfeed.tech/tags/production.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [scale](<https://devfeed.tech/tags/scale.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>), [tech](<https://devfeed.tech/tags/tech.md>)

### AI overview

Dream11 built the Darwin Feature Store as shared infrastructure for managing features consistently across offline training and real-time inference. The authors report it serves more than 200 million feature requests per minute at p99 read latency under 5 ms, with no production incidents in the prior year. The earlier setup relied on batch pipelines and Cassandra tables; it could not support real-time writes, required manual coordination for schema changes, struggled with Kafka Connect under high-frequency ingestion, and involved reactive, operationally heavy scaling. The excerpt introduces the system but does not provide its implementation details.

### Source excerpt

By Mohit Jain and Ujjwal Bagrania Dream Horizon, our open-source effort to make Dream11's battle-tested tech available to every developer, brings you the Darwin Feature Store -- a unified, low-latency feature platform built to power real-time ML at scale, and shaped to help teams build, manage, and trust features in production. Explore the Darwin Feature Store here. At Dream11, data and ML have always powered how millions of users experience sports in real time. From personalisation and relevance to ensuring efficiency under peak match traffic, ML sits deep in the critical path of the product. But models are only half the story. The real challenge is managing features -- how they're defined, versioned, and accessed consistently across training and live traffic, especially under peak sports pressure. And at the scale we operate at, feature access doesn't mean the occasional lookup; it means hundreds of millions of requests per minute, spanning both real-time inference and offline training pipelines. As Dream11 grew, feature engineering stopped being something we could manage with pipelines and tables, and became infrastructure that everything else depended on. That shift forced us to build the Darwin Feature Store: a system designed not just to tackle sports-scale traffic, but to make feature engineering predictable, reliable, and developer-friendly in production. Today, that translates into serving 200M+ feature requests per minute, delivering p99 read latencies under 5 ms, and doing so reliably through the most demanding live sports moments, without a single production incident over the last year. In this post, we'll walk through how we built the Feature Store, and what it takes to serve features reliably. The Early Days: Batch-Only, Fragile, and Manual Before the Darwin Feature Store existed, our feature engineering was mostly a few pipelines, a lot of Cassandra tables, and an increasing number of engineers building ML models. Here's what the first version looked li

## How Mixpeek Uses Qdrant for Efficient Multimodal Feature Stores

DevFeed: [How Mixpeek Uses Qdrant for Efficient Multimodal Feature Stores](<https://devfeed.tech/articles/how-mixpeek-uses-qdrant-for-efficient-multimodal-feature-stores-46618.md>)

Original publisher: [Read original article](<https://qdrant.tech/blog/case-study-mixpeek/>)

Author: info@qdrant.tech (Andrey Vasnetsov)

Published: 2025-04-08T00:00:00Z

Content type: article

Language: en

Sources: [Qdrant Blog on Qdrant - Vector Search Engine](<https://devfeed.tech/sources/qdrant-blog-on-qdrant-vector-search-engine.md>)

Topics: [Qdrant](<https://devfeed.tech/topics/qdrant.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [hybrid-search](<https://devfeed.tech/topics/hybrid-search.md>), [AI search](<https://devfeed.tech/topics/ai-search.md>), [multimodal-ai](<https://devfeed.tech/topics/multimodal-ai.md>), [MongoDB](<https://devfeed.tech/topics/mongodb.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>), [data](<https://devfeed.tech/topics/data.md>)

Tags: [approximate-nearest-neighbor-search](<https://devfeed.tech/tags/approximate-nearest-neighbor-search.md>), [bert](<https://devfeed.tech/tags/bert.md>), [data](<https://devfeed.tech/tags/data.md>), [database](<https://devfeed.tech/tags/database.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [fasttext](<https://devfeed.tech/tags/fasttext.md>), [feature](<https://devfeed.tech/tags/feature.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [hnsw](<https://devfeed.tech/tags/hnsw.md>), [hybrid-search](<https://devfeed.tech/tags/hybrid-search.md>), [image-search](<https://devfeed.tech/tags/image-search.md>), [knn-algorithm](<https://devfeed.tech/tags/knn-algorithm.md>), [matching](<https://devfeed.tech/tags/matching.md>), [mongodb](<https://devfeed.tech/tags/mongodb.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [neural-network](<https://devfeed.tech/tags/neural-network.md>), [qdrant](<https://devfeed.tech/tags/qdrant.md>), [recommender-system](<https://devfeed.tech/tags/recommender-system.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [saas](<https://devfeed.tech/tags/saas.md>), [simaes-networks](<https://devfeed.tech/tags/simaes-networks.md>), [similarity](<https://devfeed.tech/tags/similarity.md>), [transformer](<https://devfeed.tech/tags/transformer.md>), [vector-search](<https://devfeed.tech/tags/vector-search.md>), [vector-search-engine](<https://devfeed.tech/tags/vector-search-engine.md>), [vectors](<https://devfeed.tech/tags/vectors.md>), [word2vec](<https://devfeed.tech/tags/word2vec.md>)

### AI overview

This case study explains how Mixpeek uses Qdrant as a multimodal feature store for complex retrieval workflows. It describes limitations encountered with MongoDB Atlas vector search, including support for hybrid retrieval and multi-vector indexing, and reports that Qdrant simplified retriever implementations and reduced hybrid search code.

### Source excerpt

How Mixpeek Uses Qdrant for Efficient Multimodal Feature Stores About Mixpeek Mixpeek is a multimodal data processing and retrieval platform designed for developers and data teams. Founded by Ethan Steininger, a former MongoDB search specialist, Mixpeek enables efficient ingestion, feature extraction, and retrieval across diverse media types including video, images, audio, and text. The Challenge: Optimizing Feature Stores for Complex Retrievers As Mixpeek's multimodal data warehouse evolved, their feature stores needed to support increasingly complex retrieval patterns. Initially using MongoDB Atlas's vector search, they encountered limitations when implementing hybrid retrievers combining dense and sparse vectors with metadata pre-filtering.

## Introducing Bento, Snap's ML Platform

DevFeed: [Introducing Bento, Snap's ML Platform](<https://devfeed.tech/articles/introducing-bento-snap-s-ml-platform-56691.md>)

Original publisher: [Read original article](<https://eng.snap.com/introducing-bento>)

Author: Snap Platform Engineering Team

Published: 2025-01-28T20:00:00Z

Content type: article

Language: en

Sources: [Snap](<https://devfeed.tech/sources/snap.md>)

Topics: [MLOps](<https://devfeed.tech/topics/mlops.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [data](<https://devfeed.tech/topics/data.md>), [big-data](<https://devfeed.tech/topics/big-data.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [architecture](<https://devfeed.tech/tags/architecture.md>), [array](<https://devfeed.tech/tags/array.md>), [data](<https://devfeed.tech/tags/data.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-engineering](<https://devfeed.tech/tags/ml-engineering.md>), [ml-ops](<https://devfeed.tech/tags/ml-ops.md>), [ml-platform](<https://devfeed.tech/tags/ml-platform.md>), [model-deployment](<https://devfeed.tech/tags/model-deployment.md>), [model-training](<https://devfeed.tech/tags/model-training.md>), [models](<https://devfeed.tech/tags/models.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [platform](<https://devfeed.tech/tags/platform.md>), [scale](<https://devfeed.tech/tags/scale.md>)

### AI overview

Snap's Bento is an ML platform for managing server-side machine-learning workloads at scale. The article explains its architecture, end-to-end development lifecycle, feature and training data pipelines, model experimentation, deployment, and the operational challenges of running real-time ML applications in production.

### Source excerpt

Building ML at scale: Discover how Snap's Bento ML Platform powers personalized user experiences through cutting-edge ML Ops

## Declarative Feature Engineering at PayPal

DevFeed: [Declarative Feature Engineering at PayPal](<https://devfeed.tech/articles/declarative-feature-engineering-at-paypal-31933.md>)

Original publisher: [Read original article](<https://medium.com/paypal-tech/declarative-feature-engineering-at-paypal-eddcae81c06d?source=rss----6423323524ba---4>)

Author: Marina Lyan

Published: 2023-12-11T15:56:51Z

Content type: article

Language: en

Sources: [PayPal Technology](<https://devfeed.tech/sources/paypal-technology.md>)

Topics: [Feature Engineering](<https://devfeed.tech/topics/feature-engineering.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [self-service](<https://devfeed.tech/topics/self-service.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [airbnb](<https://devfeed.tech/tags/airbnb.md>), [config](<https://devfeed.tech/tags/config.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cost-analysis](<https://devfeed.tech/tags/cost-analysis.md>), [data-scientists](<https://devfeed.tech/tags/data-scientists.md>), [declarative](<https://devfeed.tech/tags/declarative.md>), [declarative-programming](<https://devfeed.tech/tags/declarative-programming.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [execution](<https://devfeed.tech/tags/execution.md>), [existing](<https://devfeed.tech/tags/existing.md>), [feature](<https://devfeed.tech/tags/feature.md>), [feature-engineering](<https://devfeed.tech/tags/feature-engineering.md>), [features](<https://devfeed.tech/tags/features.md>), [fraud](<https://devfeed.tech/tags/fraud.md>), [learning](<https://devfeed.tech/tags/learning.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml](<https://devfeed.tech/tags/ml.md>), [payment](<https://devfeed.tech/tags/payment.md>), [paypal](<https://devfeed.tech/tags/paypal.md>), [prevent-fraud](<https://devfeed.tech/tags/prevent-fraud.md>)

### AI overview

PayPal describes declarative feature engineering, which it calls config-based feature engineering, as letting data scientists specify what a feature should be while the system handles how to build it across batch, near-real-time, and real-time platforms. Its self-service UI and APIs support simulation, productization, historical backfills, and feature registration for reuse. Automation generates and optimizes pipelines, enforces point-in-time computation and engineering standards, and tracks time to market. The approach is aimed especially at simple features such as roll-up aggregations and categorical values, where standardized implementations can improve reuse, customization, cost estimation, and maintenance. More complex code-based features and analytical platforms for graphs, NLP, and anomaly detection are harder to standardize, customize, and maintain. PayPal presents the approach as effective for simple features at scale; the excerpt does not give quantitative results or the promised technical deep dive.

### Source excerpt

Photo by fabio on Unsplash PayPal supports over 400 million active consumers and merchants worldwide. Every minute there are several thousand payment transactions. To prevent fraud in real-time at such a scale, we need to streamline our ML workflow and feature engineering processes to build strong predictors of behaviors and risk indicators. On top of that, it must be done with consistently predictable Time to Market (TTM) and sustainable Total Cost of Ownership (TCO). What is declarative feature engineering? While the declarative feature engineering term was first introduced by Zipline AirBnB in 2019, we have successfully used this paradigm in PayPal for the last decade, though we know it as config-based feature engineering. The idea is to allow data scientists to write a declaration of what their features look like rather than explicitly specify how to construct them on top of different execution platforms. In this way, the concerns of feature construction and execution are abstracted away from scientists so that engineers can worry about those complexities. This post is the first in a series of two posts that will outline how the declarative feature engineering approach helps our engineers to address scale, TTM, and TCO requirements. Let's start with the definitions of metrics that measure our success. TTM of the ML Feature TTM of the Machine Learning (ML) feature is the length of time from the feature concept until it is released to production. Predictable TTM is a paramount need for overall AI maturation and business-first AI strategies. Cost of the ML Feature To qualify and estimate TCO, we break down the cost of the feature as follows: Layers of feature cost The main insight from the feature cost analysis is that we have to reuse existing features across the teams whenever possible. Otherwise, we just pay twice for every aspect. To approach the above challenges, we separate features into three levels of complexity so we can define tailored strategies for scal

## Why We Built Our Feature Store in Snowflake's Snowpark (And Moved Away From SQL)

DevFeed: [Why We Built Our Feature Store in Snowflake's Snowpark (And Moved Away From SQL)](<https://devfeed.tech/articles/why-we-built-our-feature-store-in-snowflake-s-snowpark-and-moved-away-from-sql-83131.md>)

Original publisher: [Read original article](<https://montecarlo.ai/blog-snowflake-snowpark-feature-store-architecture>)

Author: Ryan Kearns

Published: 2023-08-31T12:20:00Z

Content type: article

Language: en

Sources: [Monte Carlo](<https://devfeed.tech/sources/monte-carlo.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [Snowflake streaming data architecture](<https://devfeed.tech/topics/snowflake-streaming-data-architecture.md>), [Test-driven development](<https://devfeed.tech/topics/tdd.md>)

Tags: [data-observability](<https://devfeed.tech/tags/data-observability.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [snowflake](<https://devfeed.tech/tags/snowflake.md>), [snowpark](<https://devfeed.tech/tags/snowpark.md>), [sql](<https://devfeed.tech/tags/sql.md>)

### AI overview

The article describes why Monte Carlo moved its machine learning feature store from SQL to Python in Snowflake Snowpark as its customers, data science team, and feature needs grew. It discusses the tradeoffs between SQL and Python, choosing Snowpark over PySpark, and building a shared feature registry to improve reuse, testing, and data quality checks. The team reports that an inference process fell from one hour to 20 minutes and that the new setup enabled more granular time series anomaly detection.

### Source excerpt

Defining features in SQL worked for our feature store...until it didn't. Here is how and why we decided to take advantage of Snowpark.

## Introducing Glassdoor's ML Registry: A Centralized Artifact Management Solution

DevFeed: [Introducing Glassdoor's ML Registry: A Centralized Artifact Management Solution](<https://devfeed.tech/articles/introducing-glassdoor-s-ml-registry-a-centralized-artifact-management-solution-22612.md>)

Original publisher: [Read original article](<https://medium.com/glassdoor-engineering/introducing-glassdoors-ml-registry-a-centralized-artifact-management-solution-8bff3151cd9d?source=rss----288d984af747---4>)

Author: Rich Papalia

Published: 2023-08-30T21:20:29Z

Content type: article

Language: en

Sources: [Glassdoor Engineering](<https://devfeed.tech/sources/glassdoor-engineering.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [software-development](<https://devfeed.tech/topics/software-development.md>), [Amazon SageMaker](<https://devfeed.tech/topics/amazon-sagemaker.md>)

Tags: [artifacts](<https://devfeed.tech/tags/artifacts.md>), [aws](<https://devfeed.tech/tags/aws.md>), [configuration-management](<https://devfeed.tech/tags/configuration-management.md>), [data-pipelines](<https://devfeed.tech/tags/data-pipelines.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [glassdoor](<https://devfeed.tech/tags/glassdoor.md>), [human-in-the-loop](<https://devfeed.tech/tags/human-in-the-loop.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [machine-learning-platform](<https://devfeed.tech/tags/machine-learning-platform.md>), [manage](<https://devfeed.tech/tags/manage.md>), [metadata](<https://devfeed.tech/tags/metadata.md>), [ml](<https://devfeed.tech/tags/ml.md>), [ml-platform](<https://devfeed.tech/tags/ml-platform.md>), [model](<https://devfeed.tech/tags/model.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [platform](<https://devfeed.tech/tags/platform.md>), [recommendations](<https://devfeed.tech/tags/recommendations.md>), [registries](<https://devfeed.tech/tags/registries.md>), [registry](<https://devfeed.tech/tags/registry.md>), [robust](<https://devfeed.tech/tags/robust.md>)

### AI overview

Glassdoor describes its open-sourced ML Registry as a centralized service for managing machine learning artifacts and their metadata, including models and other artifacts. It is intended to provide a single source of truth and consistent access across teams and applications, while integrating with other tools and services. The excerpt says Glassdoor chose to build its own registry after assessing paid and open-source options that did not fully meet its requirements; it acknowledges managed services can offer out-of-the-box functionality but does not provide further comparison or implementation details.

### Source excerpt

Photo by Ryunosuke Kikuno on UnsplashGlassdoor's Machine Learning Engineering and Platform Team As part of Glassdoor's journey to becoming an ML-driven company, we've created an exceptional Machine Learning Engineering and Platform team, (see our blog about building this team here) dedicated to developing the foundational infrastructure for all of our ML initiatives. We've strategically built our ML platform by utilizing a combination of buying, building, and adopting from existing open-source solutions. This approach has allowed us to leverage the best of all worlds. Our team has had the opportunity to build numerous tools from scratch, including data pipelines, human-in-the-loop tools, and more. We've also built on top of existing tools like AWS Sagemaker to incorporate powerful in-house customizations into our feature store. We have more exciting plans in flight, such as building a new recommendations platform, further open-source contributions from our platform, and continuing to build on the foundation which we've established. Today, we'll delve into our newly open-sourced ML Registry. What is an ML Registry? The ML lifecycle encompasses far more than just model development. Once a model is built, many questions arise: Where does it reside? How do we access it? What if updates or versioning are needed? Where can we store the metadata describing the model? And what about other non-model artifacts? How can we efficiently manage all of this? Enter the ML Registry. Glassdoor's ML Registry is a centralized management service for ML artifacts and all related metadata. It serves as the single source of truth for all data pertaining to ML, enabling uniform and reliable access to this data across diverse teams and applications. It seamlessly integrates with other tools and services and provides robust, feature-rich functionality. Choosing Between Buying and Building, What Differentiates Our ML Registry? While several paid and open-source model registries already exist,

## More Design Patterns For Machine Learning Systems

DevFeed: [More Design Patterns For Machine Learning Systems](<https://devfeed.tech/articles/more-design-patterns-for-machine-learning-systems-46492.md>)

Original publisher: [Read original article](<https://eugeneyan.com//writing/more-patterns/>)

Author: Eugene Yan

Published: 2023-04-23T00:00:00Z

Content type: article

Language: en

Sources: [Eugene Yan](<https://devfeed.tech/sources/eugene-yan.md>)

Topics: [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [data-engineering](<https://devfeed.tech/topics/data-engineering.md>), [Data pipelines](<https://devfeed.tech/topics/data-pipelines.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [Apache Kafka](<https://devfeed.tech/topics/apache-kafka.md>), [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [dashboards](<https://devfeed.tech/topics/dashboards.md>)

Tags: [dashboard](<https://devfeed.tech/tags/dashboard.md>), [data](<https://devfeed.tech/tags/data.md>), [data-pipelines](<https://devfeed.tech/tags/data-pipelines.md>), [design](<https://devfeed.tech/tags/design.md>), [design-patterns](<https://devfeed.tech/tags/design-patterns.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [kafka](<https://devfeed.tech/tags/kafka.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [machinelearning](<https://devfeed.tech/tags/machinelearning.md>), [production](<https://devfeed.tech/tags/production.md>), [recsys](<https://devfeed.tech/tags/recsys.md>), [systems](<https://devfeed.tech/tags/systems.md>)

### AI overview

This article presents reusable patterns for machine learning systems, including processing and aggregating raw data once, reusing processed datasets across teams, and centralizing feature computation with feature stores. It uses clickstream data, Kafka JSON events, analytics dashboards, and session-based recommenders as examples.

### Source excerpt

9 patterns including HITL, hard mining, reframing, cascade, data flywheel, business rules layer, and more.

## Data Science Project Quick-Start

DevFeed: [Data Science Project Quick-Start](<https://devfeed.tech/articles/data-science-project-quick-start-46525.md>)

Original publisher: [Read original article](<https://eugeneyan.com//writing/project-quick-start/>)

Author: Eugene Yan

Published: 2022-03-06T00:00:00Z

Content type: tutorial

Language: en

Sources: [Eugene Yan](<https://devfeed.tech/sources/eugene-yan.md>)

Topics: [Data Science](<https://devfeed.tech/topics/data-science.md>), [Requirements](<https://devfeed.tech/topics/requirements.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>)

Tags: [data](<https://devfeed.tech/tags/data.md>), [data-science](<https://devfeed.tech/tags/data-science.md>), [datascience](<https://devfeed.tech/tags/datascience.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [how-to](<https://devfeed.tech/tags/how-to.md>), [productivity](<https://devfeed.tech/tags/productivity.md>), [reproducibility](<https://devfeed.tech/tags/reproducibility.md>), [requirements](<https://devfeed.tech/tags/requirements.md>)

### AI overview

A data science practitioner shares lessons for starting projects effectively, emphasizing business intent, context, success metrics, appropriate scoping, and reproducible engineering practices. Examples cover recommendation engines, feature stores, and search.

### Source excerpt

Hard-won lessons on how to start data science projects effectively.

## MLOps in Glassdoor: an à la carte approach

DevFeed: [MLOps in Glassdoor: an à la carte approach](<https://devfeed.tech/articles/mlops-in-glassdoor-an-a-la-carte-approach-22614.md>)

Original publisher: [Read original article](<https://medium.com/glassdoor-engineering/mlops-in-glassdoor-an-%C3%A0-la-carte-approach-b7b7703b5fd0?source=rss----288d984af747---4>)

Author: Zixin Wu

Published: 2022-01-12T00:42:43Z

Content type: article

Language: en

Sources: [Glassdoor Engineering](<https://devfeed.tech/sources/glassdoor-engineering.md>)

Topics: [MLOps](<https://devfeed.tech/topics/mlops.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [model-deployment](<https://devfeed.tech/topics/model-deployment.md>), [ML model versioning](<https://devfeed.tech/topics/ml-model-versioning.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>), [Grafana](<https://devfeed.tech/topics/grafana.md>)

Tags: [airflow](<https://devfeed.tech/tags/airflow.md>), [data-science](<https://devfeed.tech/tags/data-science.md>), [devops](<https://devfeed.tech/tags/devops.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [experimentation](<https://devfeed.tech/tags/experimentation.md>), [glassdoor](<https://devfeed.tech/tags/glassdoor.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml-models](<https://devfeed.tech/tags/ml-models.md>), [mlflow](<https://devfeed.tech/tags/mlflow.md>), [mlops](<https://devfeed.tech/tags/mlops.md>), [model-training](<https://devfeed.tech/tags/model-training.md>), [production](<https://devfeed.tech/tags/production.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>)

### AI overview

Glassdoor adopted an à la carte MLOps stack because it found no single mature solution that met its needs, and wanted component-level choice, flexibility, and faster feedback. For real-time inference, MLflow packages and serves Python models as REST services in Kubernetes, with Jenkins building images and Spinnaker deploying them; Prometheus and Grafana support metrics and alerting. Batch work uses Jenkins for simpler jobs, Airflow for multi-step workflows, and Spark when data volume or throughput calls for distributed processing. MLflow tracks experiments and registers models. Glassdoor chose not to adopt Kubeflow because its pipelines and learning curve exceeded current needs, and mostly develops models locally because remote development is not usually required. SageMaker Feature Store underlies a custom metadata and access layer; whylogs monitors data and model performance, with OpsGenie integration and threshold-based retraining described as future plans. The authors say this setup lets scientists focus on models while engineering builds supporting platforms.

### Source excerpt

Authors: Zixin Wu, Srinivasan Ramaraju Introduction Data Science and Machine Learning (ML) have become an essential toolkit for businesses to increase their competitiveness and provide greater value to their customers in recent years, largely thanks to the abundantly available large data sets, more cost-effective computation resources, and advancements in ML algorithms. In Glassdoor, we utilize ML to build intelligent products which help our job seekers find the jobs and companies they love. While ML can unlock new capabilities for businesses, it comes with its complexity and challenges compared to traditional software engineering, mainly due to data dependency and team skillset shift. Data dependency: a major reason for using ML approaches to build complex software is exactly because the desired behavior of the software is so complex that it's more efficient to be driven by data. This dependency brings new components in the software development process: Instead of simply compiling source codes, now the software (ML models) needs to be trained/retrained by (usually a large amount of) data before it can be used to serve clients Evaluation and test of ML model behaviors is data-centric and often involves offline experimentation Version tracking of ML models needs to consider both code and data changes The data could change over time due to various reasons, such as the way the data is collected or processed, or customer behavior changes which, in turn, could be driven by the effect of using the ML model. This essentially creates a hidden feedback loop. All these changes in data need to be monitored and acted accordingly. Team skillset: in many companies, the majority of the ML team members are data scientists or ML researchers, specialized in dealing with data and ML algorithms. However, to integrate an ML solution to existing systems and continuously operate it in production, many more components are needed besides ML model training and evaluation. Automating and stan

## Space efficient machine learning feature stores using probabilistic data structures - a benchmark

DevFeed: [Space efficient machine learning feature stores using probabilistic data structures - a benchmark](<https://devfeed.tech/articles/space-efficient-machine-learning-feature-stores-using-probabilistic-data-structures-a-benchmark-47432.md>)

Original publisher: [Read original article](<https://engineering.zalando.com/posts/2021/10/space-efficient-machine-learning-feature-stores-using-probabilistic-data-structures.html>)

Author: Enno Shioji

Published: 2021-10-04T22:00:00Z

Content type: article

Language: en

Sources: [Zalando](<https://devfeed.tech/sources/zalando-engineering-blog.md>)

Topics: [feature-store](<https://devfeed.tech/topics/feature-store.md>), [recommendation systems](<https://devfeed.tech/topics/recommendation-systems.md>), [bloom-filter](<https://devfeed.tech/topics/bloom-filter.md>), [Data structures](<https://devfeed.tech/topics/data-structures.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [systems](<https://devfeed.tech/topics/systems.md>)

Tags: [backend](<https://devfeed.tech/tags/backend.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [bloom-filter](<https://devfeed.tech/tags/bloom-filter.md>), [data-structures](<https://devfeed.tech/tags/data-structures.md>), [feature-store](<https://devfeed.tech/tags/feature-store.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [latency](<https://devfeed.tech/tags/latency.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [ml](<https://devfeed.tech/tags/ml.md>), [network](<https://devfeed.tech/tags/network.md>), [performance](<https://devfeed.tech/tags/performance.md>), [recommender-systems](<https://devfeed.tech/tags/recommender-systems.md>), [scalability](<https://devfeed.tech/tags/scalability.md>), [zalando](<https://devfeed.tech/tags/zalando.md>), [zalando-science](<https://devfeed.tech/tags/zalando-science.md>)

### AI overview

This post describes a space-efficient machine-learning feature store using probabilistic data structures, including a Bloom filter as the backing data store. It benchmarks the approach as an alternative to key-value-store-based feature stores in certain situations.

### Source excerpt

In this post, we describe a technique for providing machine learning feature stores with sublinear space requirements, and perform a benchmark that uses bloom filter as the backing data store. Such feature stores can be an effective alternative to the commonly used key-value-store-based feature stores in certain situations.

## Teradata Solves Biggest Challenges of Executing Analytics at Scale for Enterprise Customers

DevFeed: [Teradata Solves Biggest Challenges of Executing Analytics at Scale for Enterprise Customers](<https://devfeed.tech/articles/teradata-solves-biggest-challenges-of-executing-analytics-at-scale-for-enterprise-customers-81037.md>)

Original publisher: [Read original article](<https://www.teradata.com/press-releases/2021/teradata-solves-biggest-challenges-of-executing-analytics-at-scale-for-enterprise-customers>)

Author: January Machold

Published: 2021-09-08T07:00:00Z

Content type: release

Language: en

Sources: [Teradata News](<https://devfeed.tech/sources/teradata-news.md>)

Topics: [AI Development](<https://devfeed.tech/topics/ai-development.md>), [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Apache Spark](<https://devfeed.tech/topics/spark.md>), [feature-store](<https://devfeed.tech/topics/feature-store.md>)

Tags: [ai-and-machine-learning](<https://devfeed.tech/tags/ai-and-machine-learning.md>), [article](<https://devfeed.tech/tags/article.md>), [byom](<https://devfeed.tech/tags/byom.md>), [code](<https://devfeed.tech/tags/code.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [source](<https://devfeed.tech/tags/source.md>)

### AI overview

Teradata announced Bring Your Own Model functionality for its Vantage platform, allowing predictive models built with external and open-source tools to be imported and scored in parallel across data volumes. The release describes Analytics 1-2-3 as a process for data preparation, model training, and deployment, with AnalyticsOps monitoring model performance and triggering rescoring or updates.

### Source excerpt

New Bring Your Own Model (BYOM) functionality integrates open-source technologies, complementing an analytics framework that helps businesses operationalize artificial intelligence (AI) and machine learning (ML) projects at scale

[Next page](<https://devfeed.tech/topics/feature-store.md?cursor=WyIyMDIxLTA5LTA4VDA3OjAwOjAwKzAwOjAwIiwgIjdiMWE0MGQ2LTUzM2MtNGZhNC05ODk2LTc2MTdiMGUzMmJlNyJd>)