# Zalando Published articles for Zalando. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Scaling Reliable Experimentation in a Two-Sided AdTech Marketplace: ZMS Budget Split DevFeed: [Scaling Reliable Experimentation in a Two-Sided AdTech Marketplace: ZMS Budget Split]() Original publisher: [Read original article]() Author: Stefano Restelli Published: 2026-08-31T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [experiments](), [Concurrency](), [Scalability]() Tags: [adtech](), [backend](), [bias](), [concurrency](), [data](), [data-science](), [experimentation](), [experiments](), [machine-learning](), [scaling](), [zalando]() ### AI overview The article explains how a two-sided retail media marketplace can invalidate conventional A/B tests when experiment variants draw from shared advertiser budgets. It describes Budget Split and Orthogonal Concurrency as mechanisms for reducing budget cannibalization bias and supporting reliable concurrent experimentation. ### Source excerpt Budget Split and Orthogonal Concurrency reduce budget cannibalization bias and enable reliable concurrent experiments in two-sided AdTech marketplaces. ## Agentic Engineering at Zalando: a snapshot DevFeed: [Agentic Engineering at Zalando: a snapshot]() Original publisher: [Read original article]() Author: Bartosz Ocytko Published: 2026-08-13T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [agentic-coding](), [Large Language Model](), [API](), [litellm](), [Cache](), [Command-line interface](), [GitHub Copilot](), [Pydantic](), [Memory Leaks]() Tags: [agentic](), [agentic-engineering](), [ai](), [api](), [artificial-intelligence](), [caching](), [cli](), [culture](), [engineering](), [hackathon](), [ide](), [litellm](), [llm](), [llms](), [machine-learning](), [memory-leak](), [open-source](), [zalando]() ### AI overview Zalando shares lessons from 2.5 years of agentic engineering across more than 250 engineering teams. The article describes a LiteLLM-based proxy for accessing multiple LLM providers, adoption and cost tracking, prompt caching, model and client controls, a chat UI, and a custom CLI. ### Source excerpt We look back at our journey of agentic engineering at Zalando, sharing our learnings and approaches that worked well for us in the past 2.5 years. ## From Homegrown to Flink: Migrating a Stateful Ad Event Join at Scale DevFeed: [From Homegrown to Flink: Migrating a Stateful Ad Event Join at Scale]() Original publisher: [Read original article]() Author: Aleksandr Birin Published: 2026-07-23T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Apache Flink](), [stream-processing](), [Data pipelines](), [JOIN](), [real-time](), [Stateful](), [Apache Kafka](), [Scalability](), [Java]() Tags: [apache-flink](), [backend](), [data-pipelines](), [distributed](), [event](), [flink](), [java](), [kafka](), [scale](), [stateful](), [zalando]() ### AI overview Zalando's Ad Platform team migrated a homegrown Java application that joined bid and interaction events from Kinesis and Nakadi streams to Apache Flink. The migration addressed partitioning and state-loss problems while halving infrastructure cost and eliminating in-memory state loss on restarts. ### Source excerpt We replaced a homegrown in-memory stream join with Apache Flink, halved our infrastructure cost, and eliminated in-memory state loss on restarts. ## Client-Side Load Balancing at a Million Requests Per Second DevFeed: [Client-Side Load Balancing at a Million Requests Per Second]() Original publisher: [Read original article]() Author: Conor Gallagher Published: 2026-06-22T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Load Balancing](), [Latency](), [Cache](), [Kubernetes](), [Availability](), [Scalability](), [Network]() Tags: [availability](), [backend](), [cache](), [client](), [cluster](), [infrastructure](), [kubernetes](), [latency](), [load-balancer](), [load-balancing](), [low-latency](), [open-source](), [routing](), [scalability](), [scaling](), [source](), [sre](), [traffic](), [zalando]() ### AI overview Zalando describes replacing an internal shared-ingress routing path with an in-process client-side load balancer for high-volume fan-out traffic. The approach adds N-ring fade-in, occupancy-based bounded load, and availability-zone-aware routing with a latency health factor, improving observability, resilience, and cost efficiency. ### Source excerpt How we built an in-process client-side load balancer for a million requests per second of internal fan-out traffic, what we layered on top (N-ring fade-in, occupancy-based bounded load, and AZ-aware routing with a latency health factor), and how hardening that path cut cost and made the service resilient to the infrastructure underneath it. ## Introducing Lightstep UQL to PromQL Translator DevFeed: [Introducing Lightstep UQL to PromQL Translator]() Original publisher: [Read original article]() Author: Maksim Pershin Published: 2026-06-07T22:00:00Z Content type: tutorial Language: en Sources: [Zalando]() Topics: [migration](), [telemetry](), [Go Language](), [SDK](), [Parsing](), [Automation](), [REST API]() Tags: [api](), [automation](), [backend](), [go](), [golang](), [migration](), [open-source](), [parsing](), [rest-api](), [sdk](), [sre](), [telemetry](), [zalando]() ### AI overview Zalando describes an open-source Go translator that converts Lightstep UQL telemetry queries into PromQL to support migration from Lightstep to Dash0. It is available as an SDK, HTTP server, web UI, and REST API. ### Source excerpt Automate telemetry query migration from Lightstep UQL to PromQL with this open-source Go SDK and Web UI. ## Rejecting Invalid Ingress Routes at Apply Time DevFeed: [Rejecting Invalid Ingress Routes at Apply Time]() Original publisher: [Read original article]() Author: Veronika Volokitina Published: 2026-04-08T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Kubernetes](), [control-plane](), [proxy](), [Routing (disambiguation)](), [api server](), [configuration](), [Parsing]() Tags: [api](), [api-server](), [backend](), [backends](), [configuration](), [control-plane](), [kubernetes](), [open-source](), [parsing](), [platform-engineering](), [reverse-proxy](), [risk](), [routing](), [scale](), [server](), [skipper](), [sre](), [validating](), [webhook](), [zalando]() ### AI overview Zalando describes using Skipper as a Kubernetes validating admission webhook to reject invalid ingress routes before they are applied. The approach reuses Skipper's runtime validation for filters and predicates, addressing configuration errors that Kubernetes can accept as structurally valid. ### Source excerpt How Zalando used Skipper as a validating admission webhook to reject invalid filters and predicates at apply time, and what it took to make that safe on the Kubernetes control-plane path. ## How Zalando Uses LLM-as-a-Judge for Search Quality Assurance DevFeed: [How Zalando Uses LLM-as-a-Judge for Search Quality Assurance]() Original publisher: [Read original article]() Author: Tao Ruangyam Published: 2026-03-16T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Large Language Model](), [Artificial Intelligence](), [test-coverage](), [data]() Tags: [ai](), [airflow](), [artificial-intelligence](), [backend](), [coverage](), [developers](), [evaluation](), [llm](), [llm-as-a-judge](), [machine-learning](), [search](), [zalando]() ### AI overview Zalando describes a framework that uses LLM-as-a-judge techniques to evaluate search-result relevance at scale, with broad test coverage and multilingual support. The approach is intended to identify and diagnose search issues proactively before launching in new markets. ### Source excerpt Deep dive into how Zalando builds a search quality assurance framework with LLM-as-a-judge to evaluate the search quality at scale with high coverage and multi-language support. ## Migrating jackson-datatype-money to FasterXML: A Case Study in Open Source Consolidation DevFeed: [Migrating jackson-datatype-money to FasterXML: A Case Study in Open Source Consolidation]() Original publisher: [Read original article]() Author: Sri Adarsh Kumar Published: 2026-03-08T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Jackson](), [LocalDate](), [Open Source](), [migration](), [Java](), [Library](), [Maintainers](), [Maven]() Tags: [dependency](), [jackson](), [java](), [libraries](), [library](), [maintainers](), [maven](), [migration](), [open-source](), [zalando]() ### AI overview A case study of migrating Zalando's jackson-datatype-money module into the official FasterXML Jackson ecosystem. It explains the fragmented support for JSR 354 monetary types, the consolidation rationale, and architectural changes including separate generic and Moneta-specific modules. ### Source excerpt How we integrated jackson-datatype-money into the official FasterXML Jackson ecosystem. ## Why We Ditched Flink Table API Joins: Cutting State by 75% with DataStream Unions DevFeed: [Why We Ditched Flink Table API Joins: Cutting State by 75% with DataStream Unions]() Original publisher: [Read original article]() Author: Maryna Kryvko Published: 2026-03-03T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Apache Flink](), [stream-processing](), [SQL](), [Stateful](), [rocksdb](), [Amazon Web Services]() Tags: [apache-flink](), [aws](), [backend](), [clusters](), [flink](), [memory](), [search](), [sql](), [state](), [stream-processing](), [streaming-data](), [zalando]() ### AI overview This article explains why Zalando replaced Apache Flink Table API and SQL joins with a manually controlled DataStream API implementation for a product-enrichment pipeline. It discusses state accumulation in chained joins, RocksDB-backed state, and the reported reduction from 240GB to 56GB. ### Source excerpt The beauty of a high-level abstraction is that it lets you focus on the "what" rather than the "how." In the world of Apache Flink, the Table API is a powerful tool that abstracts away the complexities of stream processing, allowing developers to write SQL-like queries on streaming data. However, as we discovered in our journey with Flink, there are scenarios where the Table API's abstraction can be too heavy. ## Running an Engineering Papers Reading Guild at Zalando DevFeed: [Running an Engineering Papers Reading Guild at Zalando]() Original publisher: [Read original article]() Author: Danilo Veljovic Published: 2026-01-28T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Learning](), [meetings](), [Computing](), [Software](), [distributed-systems]() Tags: [computing](), [culture](), [distributed-systems](), [engineering](), [guild](), [meetings](), [papers](), [reading](), [zalando]() ### AI overview Zalando's Engineering Papers guild reflects on its first year of monthly meetups for reading and discussing academic papers. The article explains the group's motivations, including understanding the foundations beneath software abstractions, and shares lessons for organising a sustainable internal paper-reading community. ### Source excerpt In September 2024, we started an Engineering Papers guild at Zalando to read and discuss research papers together. A year later, we reflect on our journey and share insights on organising and evolving the guild. ## Paper Announcement: A Practical Approach to Replenishment Optimization with Extended (R, s, Q) Policy and Probabilistic Models DevFeed: [Paper Announcement: A Practical Approach to Replenishment Optimization with Extended (R, s, Q) Policy and Probabilistic Models]() Original publisher: [Read original article]() Author: Alva Presbitero Published: 2026-01-14T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Optimization](), [Simulation](), [data](), [Risk]() Tags: [artificial-intelligence](), [backend](), [data](), [demand-forecasting](), [logistics](), [machine-learning](), [models](), [monte-carlo](), [operations-research](), [optimization](), [paper](), [probabilistic](), [risk](), [simulation](), [zalando](), [zalando-science]() ### AI overview Zalando describes a replenishment optimization engine that combines probabilistic demand forecasting, discrete event simulation, and stochastic optimization. Its extended (R, s, Q) policy adapts replenishment across a product's lifecycle while Monte Carlo simulation evaluates candidate policies under uncertainty. ### Source excerpt Learn how the ZEOS replenishment optimization system achieves up to 22.1% GMV uplift by unifying probabilistic demand forecasting with risk-aware discrete event simulation. ## Contributing to Debezium: Fixing Logical Replication at Scale DevFeed: [Contributing to Debezium: Fixing Logical Replication at Scale]() Original publisher: [Read original article]() Author: Conor Gallagher Published: 2025-12-18T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Replication](), [Event-Streaming](), [PostgreSQL](), [Streaming](), [Database](), [Kubernetes]() Tags: [backend](), [databases](), [debezium](), [event-streaming](), [infrastructure](), [kubernetes](), [kubernetes-clusters](), [open-source](), [postgresql](), [replication](), [scale](), [streaming](), [wal](), [zalando]() ### AI overview Zalando describes how its engineers contributed two Debezium features to address PostgreSQL logical replication problems in a Kubernetes-based event-streaming platform. The article focuses on runaway WAL growth in low-activity databases and a compatibility issue involving pgjdbc keepalive flushing during a Debezium upgrade. ### Source excerpt How we contributed two features to Debezium to solve WAL growth and enable safer logical replication for our event streaming infrastructure ## The Day Our Own Queries DoS'ed Us: Inside Zalando Search DevFeed: [The Day Our Own Queries DoS'ed Us: Inside Zalando Search]() Original publisher: [Read original article]() Author: Maryna Kryvko Published: 2025-12-16T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [elasticsearch](), [incident](), [Back end](), [Query (disambiguation)](), [systems](), [AI search](), [real-time]() Tags: [backend](), [backends](), [dos](), [elasticsearch](), [full-text-search](), [incident](), [optimizing](), [query](), [real-time](), [search](), [sre](), [zalando]() ### AI overview Zalando's Search & Browse team recounts an incident in which an Elasticsearch cluster became sluggish and unresponsive because of a self-inflicted denial-of-service attack. The article describes the customer, partner, and campaign impact of degraded catalog search and outlines the system layers involved under load, including lexical and vector search, query building, and named entity recognition. ### Source excerpt Once upon a time, during a normal Sunday, our team ran into an unexpected challenge: an Elasticsearch cluster that suddenly became sluggish and unresponsive due to a self-inflicted Denial of Service (DoS) attack (of course we didn't know it at the time). This is the story of how we identified, mitigated, and learned from this incident. ## Accelerating Mobile App development at Zalando with Rendering Engine and React Native DevFeed: [Accelerating Mobile App development at Zalando with Rendering Engine and React Native]() Original publisher: [Read original article]() Author: Rene Eichhorn Published: 2025-10-02T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Mobile](), [React Native](), [cross-platform](), [migration](), [User Interfaces](), [interoperability](), [navigation](), [animations]() Tags: [animations](), [cross-platform](), [frontend](), [interoperability](), [ios](), [migration](), [mobile](), [open-source](), [react](), [react-native](), [ui](), [zalando](), [zalando-app]() ### AI overview Zalando describes its gradual migration from separate iOS and Android codebases to a cross-platform mobile architecture using React Native and its internal Rendering Engine. The article explains the decision criteria, including faster delivery, progressive adoption, shared web and mobile capabilities, and native interoperability. ### Source excerpt We present how we combined our internal React-based UI composition framework Rendering Engine with React Native in a brownfield integration approach that enables us to gradually modernise our mobile app technology stack while at the same time introducing cross-platform technologies to build shared experiences. ## Dead Ends or Data Goldmines? Investment Insights from Two Years of AI-Powered Postmortem Analysis DevFeed: [Dead Ends or Data Goldmines? Investment Insights from Two Years of AI-Powered Postmortem Analysis]() Original publisher: [Read original article]() Author: Dmitry Kolesnikov Published: 2025-09-24T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [site-reliability-engineering](), [Artificial Intelligence](), [incident](), [Scalability](), [trust](), [Amazon DynamoDB](), [Amazon S3](), [elasticsearch]() Tags: [ai](), [artificial-intelligence](), [automation](), [aws-s3](), [backend](), [datastores](), [decision-making](), [dynamodb](), [elasticsearch](), [hallucinations](), [infrastructure](), [llm](), [machine-learning](), [postmortem](), [reliability](), [reliability-engineering](), [scaling](), [sre](), [zalando]() ### AI overview Zalando describes using large language models as SRE assistants to analyze years of postmortems across datastore technologies. The approach identifies recurring incident patterns, systemic hotspots, and infrastructure investment opportunities, while human curation remains necessary to address hallucinations and attribution errors. ### Source excerpt Your incidents hold the blueprint to your most strategic infrastructure wins -- if you're listening correctly. ## Direct Data Sharing using Delta Sharing - Introduction: Our Journey to Empower Partners at Zalando DevFeed: [Direct Data Sharing using Delta Sharing - Introduction: Our Journey to Empower Partners at Zalando]() Original publisher: [Read original article]() Author: Lokeshbabu Radhakrishnan Published: 2025-07-07T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [data](), [data-processing]() Tags: [big-data](), [data](), [pilot](), [platform](), [processing](), [program](), [real-time](), [sharing](), [zalando]() ### AI overview Zalando describes how its Partner Tech division addressed fragmented partner data access by piloting and scaling Delta Sharing into an organization-wide platform. The post covers partner needs, the solution-selection journey, implementation lessons, and the resulting approach to secure data sharing. ### Source excerpt In this post, we explain how we transformed fragmented partner data sharing at Zalando by implementing Delta Sharing, evolving from a pilot solution to an organization-wide platform that enables real-time, secure data access across our partner ecosystem. ## Building a Dynamic Inventory Optimization System DevFeed: [Building a Dynamic Inventory Optimization System]() Original publisher: [Read original article]() Author: Alva Presbitero Published: 2025-06-29T22:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Optimization](), [Artificial Intelligence](), [decision-making](), [Simulation](), [recommendation systems]() Tags: [ai](), [applied-science](), [artificial-intelligence](), [aws](), [backend](), [building](), [cost-optimisation](), [data-science](), [deep-dive](), [demand-forecasting](), [e-commerce](), [lead-time](), [machine-learning](), [monte-carlo](), [operations-research](), [optimization](), [pipelines](), [probabilistic](), [recommendations](), [scale](), [zalando](), [zalando-science]() ### AI overview This technical article explains a dynamic inventory optimization system for e-commerce replenishment. It covers probabilistic demand and lead-time forecasting, cost optimization, recommendation generation, Monte Carlo simulation, and scalable forecasting and policy optimization pipelines. ### Source excerpt This technical blog outlines how we built a scalable inventory optimization system to help partners maintain a profitable inventory. ## Adapting to Change: Returning to Work in a Fast-Moving Tech World DevFeed: [Adapting to Change: Returning to Work in a Fast-Moving Tech World]() Original publisher: [Read original article]() Author: Kanupriya Gupta Published: 2025-05-18T22:00:00Z Content type: opinion Language: en Sources: [Zalando]() Topics: [software-development]() Tags: [career](), [changes](), [culture](), [development](), [development-environment](), [team](), [zalando]() ### AI overview An employee describes returning to work after a four-month career break and adapting to changes in team structure, technology, priorities, relationships, and development-environment responsibilities at Zalando. ### Source excerpt This article discusses the challenges of returning to work in a fast-paced tech environment and how to navigate them. ## How Zalando Replaced an Event-Driven Product Data System with a High-Performance API DevFeed: [How Zalando Replaced an Event-Driven Product Data System with a High-Performance API]() Original publisher: [Read original article]() Author: Conor Gallagher Published: 2025-03-06T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [API](), [event driven](), [Latency](), [Microservices](), [software-architecture](), [Scalability](), [Cache](), [GraphQL]() Tags: [api](), [apis](), [architecture](), [availability](), [backend](), [caching](), [event-driven](), [event-driven-architecture](), [graphql](), [latency](), [microservices](), [microservices-architecture](), [performance](), [scalability](), [scaling](), [zalando]() ### AI overview Zalando describes redesigning its event-driven product data architecture after delays and scaling challenges affected price and stock updates. The post explains the move to a high-performance serving API, including caching, low-latency optimizations, and architectural decisions intended to support millions of requests per second. ### Source excerpt In this post, we explain how we replaced an event-driven system with a high performance API capable of serving millions of requests per second with single-digit-millisecond latency at P99 ## LLM powered migration of UI component libraries DevFeed: [LLM powered migration of UI component libraries]() Original publisher: [Read original article]() Author: Naval Singh Published: 2025-02-19T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [migration](), [Large Language Model](), [User Interfaces](), [software-development](), [Hackathon]() Tags: [component](), [developers](), [development](), [frontend](), [llm](), [machine-learning](), [migration](), [ui](), [zalando]() ### AI overview Zalando describes using large language models to help migrate partner-facing applications from one in-house UI component library to another. The project covered 15 B2B applications and evaluated LLM-based automation alongside traditional JavaScript codemods, with attention to accuracy, custom components, and subtle bugs. ### Source excerpt Sharing our approach and insights employing LLMs to migrate in-house UI component libraries at scale. ## Scaling Beyond Limits: Harnessing Route Server for a Stable Cluster DevFeed: [Scaling Beyond Limits: Harnessing Route Server for a Stable Cluster]() Original publisher: [Read original article]() Author: Mustafa Abdelrahman Published: 2025-02-16T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Kubernetes](), [Scalability](), [proxy](), [control-plane](), [etcd](), [api server](), [Cache](), [reliability-engineering](), [Parsing](), [TLS (Transport Layer Security)]() Tags: [api](), [api-server](), [aws](), [backend](), [cache](), [cluster](), [control-plane](), [etcd](), [golang](), [ingress](), [kubernetes](), [kubernetes-clusters](), [open-source](), [parsing](), [performance](), [platform-engineering](), [processing](), [proxy](), [reliability](), [routing](), [scaling](), [skipper](), [tls](), [zalando]() ### AI overview Zalando describes how a custom Route Server proxy reduced Kubernetes control-plane load from its Skipper ingress controller. The design uses polling, parsing, and an HTTP ETag cache layer between Skipper and the Kubernetes API Server to improve cluster stability at large scale. ### Source excerpt A proxy server contributing to a stable Kubernetes cluster and scaling ingress controller. ## JSON Web Keys (JWK): Rotating Cryptographic Keys at Zalando DevFeed: [JSON Web Keys (JWK): Rotating Cryptographic Keys at Zalando]() Original publisher: [Read original article]() Author: Jan Brennenstuhl Published: 2025-01-20T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [Security](), [secret rotation](), [OpenID connect (OIDC)](), [JSON Web Tokens](), [Authentication](), [Authorization](), [Encryption](), [public key](), [sensitive data]() Tags: [api-keys](), [authentication](), [authorization](), [backend](), [encryption](), [jwks](), [oidc](), [openid-connect](), [public-key](), [secret-rotation](), [security](), [sensitive-data](), [zalando]() ### AI overview Zalando describes automating JSON Web Key rotation for its OpenID Connect identity provider. The article explains how JWKs distribute public keys for JWT verification and why regularly replacing cryptographic keys reduces the impact of key compromise and forged tokens. ### Source excerpt Secret rotation is a vital security measure in many contexts. Learn how we automate this process using JSON Web Keys (JWKs) to enhance the security of our customer identity provider. ## Introducing Lightstep Receiver for OpenTelemetry Collector DevFeed: [Introducing Lightstep Receiver for OpenTelemetry Collector]() Original publisher: [Read original article]() Author: Konstantin Zhukov Published: 2025-01-20T23:00:00Z Content type: release Language: en Sources: [Zalando]() Topics: [OpenTelemetry](), [tracing](), [telemetry](), [legacy systems](), [protocols](), [Kubernetes](), [gRPC](), [Go Language](), [JavaScript](), [Python]() Tags: [backend](), [cluster](), [collector](), [go](), [grpc](), [http](), [javascript](), [json](), [kubernetes](), [legacy](), [open-source](), [opentelemetry](), [python](), [sre](), [traces](), [zalando]() ### AI overview Zalando introduces a Lightstep Receiver for the OpenTelemetry Collector. The component receives traces from legacy Lightstep tracers, converts them to OpenTelemetry, and forwards them through the Collector pipeline to supported backends. The article also outlines custom Collector builds, deployment options, receiver configuration, and exporter setup. ### Source excerpt OpenTelemetry Lightstep Receiver helps you ingest traces generated by legacy Lightstep tracers in a simple way. ## Exploring the Potential of Graph Neural Networks to Transform Recommendations at Zalando DevFeed: [Exploring the Potential of Graph Neural Networks to Transform Recommendations at Zalando]() Original publisher: [Read original article]() Author: Mariia Bulycheva Published: 2024-12-18T23:00:00Z Content type: article Language: en Sources: [Zalando]() Topics: [recommendation systems](), [Graphs](), [Neural Network](), [Embeddings]() Tags: [customers](), [deep-learning](), [embeddings](), [graph](), [machine-learning](), [neural-networks](), [recommendations](), [recommender-systems](), [research](), [zalando](), [zalando-science]() ### AI overview Zalando explores integrating graph neural networks into its homepage recommender system. The approach models users and content as a bipartite graph, predicts future interactions, and generates embeddings that can be added to the existing production model. ### Source excerpt Delivering personalized recommendations is key to engaging Zalando's customers, but traditional models can miss the complexity of user-content interactions. By integrating graph neural networks (GNNs), we're exploring a cutting-edge approach to better predict clicks and enhance the shopping experience. [Next page]()