# ml Published articles for ml. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## ADLX 2.0: Extending graphics control to AI agents and agentic apps DevFeed: [ADLX 2.0: Extending graphics control to AI agents and agentic apps]() Original publisher: [Read original article]() Author: Pete Vagiakos; Alexander Blake-Davies Published: 2026-09-16T10:00:00Z Content type: article Language: en Sources: [AMD GPUOpen]() Topics: [GPU](), [Artificial Intelligence](), [Hardware](), [SDK](), [Model Context Protocol]() Tags: [agents](), [ai](), [amd-device-library-extra](), [amd-device-library-extra-adlx](), [announcement](), [developers](), [gpu](), [hardware](), [mcp](), [ml](), [product-blogs](), [product-release](), [tools]() ### AI overview AMD ADLX 2.0 adds an AI extension framework and MCP servers that connect AI applications with AMD graphics technologies, enabling developers to build apps that monitor, manage, and optimize AMD graphics hardware. ### Source excerpt AMD ADLX 2.0 adds AI extensions and MCP servers to help developers build intelligent apps that can monitor, manage, and optimize AMD graphics hardware. ## How energy teams turn theft detection into governed action with Genie and AI business processes DevFeed: [How energy teams turn theft detection into governed action with Genie and AI business processes]() Original publisher: [Read original article]() Author: Daniel Zoccali; Jack Yallop Published: 2026-09-15T16:50:00Z Content type: article Language: en Sources: [Databricks]() Topics: [databricks](), [Machine learning](), [Artificial Intelligence]() Tags: [databricks](), [energy](), [fraud](), [industries](), [ml](), [model](), [reporting](), [safety](), [speed]() ### AI overview The article explains how energy teams can operationalize energy-theft detection by connecting model-generated risk signals with investigation, field operations, revenue recovery, and reporting in a governed workflow. It presents a Databricks implementation using a Databricks App, Lakebase, and Unity Catalog. ### Source excerpt Energy theft is the deliberate use of gas or electricity without paying for it, typically... ## 🍔🧠 Pinterest's Fix for the Hardest Problem in ML Infra DevFeed: [🍔🧠 Pinterest's Fix for the Hardest Problem in ML Infra]() Original publisher: [Read original article]() Author: Alexandre Zajac Published: 2026-09-14T15:31:30Z Content type: article Language: en Sources: [Hungry Minds]() Topics: [recommendation systems](), [Feature Engineering](), [Inference](), [Streaming](), [configuration](), [Framework](), [data](), [Low Latency]() Tags: [batch](), [configuration](), [data](), [feature-engineering](), [low-latency](), [ml](), [pinterest](), [ranking](), [recommendation-systems](), [streaming]() ### AI overview Pinterest redesigned its user-sequence platform for ranking, retrieval, and recommendation systems by defining signals once and instantiating them consistently across streaming, batch, and serving workloads. The approach uses Python configuration with validated schemas, a shared execution engine, cooperating streaming and batch paths, and columnar time-partitioned storage to improve freshness, completeness, consistency, and operational efficiency. ### Source excerpt PLUS: OpenAI agents beat math 🧮, Test techniques for agents ⚡, Postgres survival guide 📖 ## Открываем претрейн Alice AI Search: как устроена модель быстрых ответов Алисы на Поиске DevFeed: [Открываем претрейн Alice AI Search: как устроена модель быстрых ответов Алисы на Поиске]() Original publisher: [Read original article]() Author: pet67 (Яндекс) Published: 2026-09-11T06:05:13Z Content type: article Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [Artificial Intelligence](), [AI search](), [Transformers](), [Mixture of Experts (MoE)](), [Large Language Model](), [hugging face]() Tags: [agentic](), [ai](), [ai-search](), [alice-ai](), [hugging-face](), [llm](), [ml](), [moe](), [rl](), [tag-178bc8f01f24](), [tag-4004cf5948d3](), [tag-61cd5a476b1d](), [tag-d89cae10e887](), [tag-e6d9cc1f0757](), [transformers]() ### AI overview This developer article explains the Alice AI Search pipeline for generating fast answers, including its search and context-processing stages, shorter information contexts, a sparse Mixture-of-Experts architecture combined with an Encoder-Decoder, and online reinforcement learning from user behavior signals. It also announces the open release of the Alice AI-T5-35B-A0.6B Base model, with external inference available through Hugging Face Transformers while optimized production inference remains internal to Yandex. ### Source excerpt Быстрый ответ Алисы AI -- это самый массовый генеративный продукт Яндекса и первое соприкосновение с Алисой для пользователей Поиска. Даже в час пиковой нагрузки пользователь должен получить лаконичный ответ за считаные секунды. Для этого мы, команда Alice AI Search, адаптируем весь пайплайн быстрых ответов -- от собственного претрейна с кастомной архитектурой до онлайн-rl-обучения на поведенческие сигналы пользователей. В статье разберём, как устроен генеративный ответ в Поиске, и расскажем про основные улучшения июньского релиза: как мы ускорили ответы за счёт коротких инфоконтекстов, зачем совместили Encoder-Decoder с разреженной MoE-архитектурой и как обучение на реальных пользовательских сигналах повлияло на качество и использование продукта. Кроме того, мы выложили в открытый доступ обученную с нуля модель Alice AI-T5-35B-A0.6B Base с тем ограничением, что внешним пользователям доступен инференс через Hugging Face Transformers, а оптимизированный production-инференс пока доступен только внутри Яндекса. Читать далее ## Creating an AI Platform for classic ML online inference DevFeed: [Creating an AI Platform for classic ML online inference]() Original publisher: [Read original article]() Author: Rohith Leeladharan Published: 2026-09-10T07:26:46Z Content type: tutorial Language: en Sources: [Amex GBT Technology]() Topics: [AI Platform](), [Inference](), [MLOps](), [Machine learning](), [model-deployment]() Tags: [ai](), [ai-platform](), [ai-platform-engineering](), [deploy](), [experiments](), [feature-store](), [inference](), [machine-learning](), [ml](), [mlops](), [monitoring](), [predictions]() ### AI overview This article describes how American Express Global Business Travel built an AI platform for deploying classic machine-learning systems and supporting online inference. It explains the platform's requirements--simplicity, self-service, experimentation, and continuous improvement--and details the pre-process, predict, post-process pattern used by inference engines. ### 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 ## Temporally stable generative illumination with a one-step diffusion model DevFeed: [Temporally stable generative illumination with a one-step diffusion model]() Original publisher: [Read original article]() Author: SungYe Kim; Harish Anand; Alexandr Kuznetsov; Wojciech Uss; Wojciech Kaliński; Rama Harihara Published: 2026-09-09T13:00:00Z Content type: article Language: en Sources: [AMD GPUOpen]() Topics: [real-time rendering](), [VAE]() Tags: [arr-group](), [article-release](), [diffusion](), [generation](), [gi](), [inference](), [lighting](), [ml](), [model](), [optimization](), [quality](), [ray-tracing](), [raytracing](), [real-time](), [real-time-rendering](), [research](), [white-paper]() ### AI overview The article presents a single-step latent diffusion method for real-time global illumination. It conditions image generation on scene signals and lighting hints, and uses a Temporal VAE decoder with motion-vector reprojection to improve temporal stability and reduce flicker. ### Source excerpt A generative method for real-time global illumination using a single-step latent diffusion model, delivering stable, high-quality lighting without costly iterative processing. ## Momentum in ML, Explained Visually and Intuitively! DevFeed: [Momentum in ML, Explained Visually and Intuitively!]() Original publisher: [Read original article]() Author: Avi Chawla Published: 2026-09-08T21:24:18Z Content type: article Language: en Sources: [Daily Dose of Data Science]() Topics: [Machine learning](), [Optimization]() Tags: [ml](), [optimization]() ### AI overview This article explains momentum in machine learning visually and intuitively, presenting it as an optimization technique for speeding model training. The supplied excerpt also previews related coverage of distributed training and hyperparameter optimization. ### Source excerpt (a popular ML interview question) ## Updates on HEIR, the homomorphic encryption compiler project DevFeed: [Updates on HEIR, the homomorphic encryption compiler project]() Original publisher: [Read original article]() Published: 2026-09-04T18:53:40Z Content type: article Language: en Sources: [Jeremy Kun]() Topics: [homomorphic encryption](), [Compiler](), [Encryption](), [Cryptography](), [Machine learning](), [Inference](), [GitHub](), [bazel](), [Kaggle]() Tags: [bazel](), [ckks](), [compiler](), [cryptography](), [encryption](), [github](), [homomorphic-encryption](), [inference](), [kaggle](), [mathematics](), [ml](), [programming]() ### AI overview This companion article explains HEIR, a homomorphic encryption compiler that converts programs to operate directly on encrypted data. It discusses compiling pre-trained machine-learning models for private inference, describes the repository and setup, and reports an example involving encrypted credit-card fraud detection. ### Source excerpt On 2026-08-14 I published an article on the Google Security blog with an update on HEIR, our homomorphic encryption (HE) compiler. This is a companion article, in which I have no limits on word count or jargon, and I can feel free to be honest. So strap in. Assuming you won't read the linked corporate blog post, HEIR is a compiler that converts an input program to a program that operates directly on encrypted data. ## Training Yandex's Alice Omnimodel to Integrate Text and Images DevFeed: [Training Yandex's Alice Omnimodel to Integrate Text and Images]() Original publisher: [Read original article]() Author: formica\_rufa (Яндекс) Published: 2026-09-03T07:03:43Z Content type: tutorial Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [Large Language Model](), [vlm](), [Artificial Intelligence](), [Mixture of Experts (MoE)](), [qwen](), [Mercury]() Tags: [ai](), [llm](), [ml](), [moe](), [qwen](), [tag-055aee430837](), [tag-61cd5a476b1d](), [tag-831b63de9433](), [tag-86b843454893](), [tag-95a2c958e46b](), [tag-ef0b1bf200df](), [vlm]() ### AI overview Yandex describes its work on an Alice omnimodel that combines a text LLM and a visual VLM into one model for text and image interactions. The article focuses on lessons from training and alignment, including the role of MoE architecture and reinforcement learning. ### Source excerpt Ещё недавно Алиса отвечала на текст и на картинку будто двумя разными голосами. Под капотом и правда жили две генеративные модели: текстовая LLM и визуальная VLM, а между ними -- стена из непрозрачного роутинга, разных форматов ответов и разной вёрстки. Почти год мы сводили их в одну омнимодель -- такую, которая воспринимает текст и изображения как единое целое, без переключений за кадром. Получилось не всё и не сразу, но путь вышел поучительным, и в этой статье я хочу поделиться тем, что мы поняли про обучение таких моделей. Попутно -- несколько неочевидных поворотов: почему за два года до этого та же затея разваливалась, что изменила MoE-архитектура, почему омнипретрейн пришлось собирать с конца и почему один вид RL переезжает на большую модель легко, а другой рассыпается прямо на глазах. Меня зовут Алексей Григорьев, я представляю большую команду разработки омнимодели Яндекса. Вместе с моим коллегой Данилой Кашиным я расскажу про все технические грабли не со стороны наблюдателя, а как их непосредственный собиратель. Но рассказывать я буду с акцентом не на красивом замысле, а на самой болезненной части -- алайнменте. Читать далее ## AMD FSR plugin updated for Unreal Engine 5.8 DevFeed: [AMD FSR plugin updated for Unreal Engine 5.8]() Original publisher: [Read original article]() Author: Joe Rozek; Alexander Blake-Davies Published: 2026-08-27T12:00:00Z Content type: release Language: en Sources: [AMD GPUOpen]() Topics: [Unreal Engine](), [releases](), [GPU](), [Machine learning](), [SDKs]() Tags: [amd-fsr-3](), [amd-fsr-4](), [amd-fsr-frame-generation](), [amd-fsr-framegeneration](), [amd-fsr-upscaling](), [announcement](), [antilag](), [fsr-3](), [fsr-4](), [game-development](), [generation](), [gpu](), [ml](), [news](), [performance](), [plugin](), [product-blogs](), [product-release](), [redstone](), [release](), [super-resolution](), [technical-articles](), [unreal](), [unreal-engine]() ### AI overview The AMD FSR Unreal Engine plugin has been updated for Unreal Engine 5.8. It adds FSR Redstone SDK 2.3 updates, FSR Upscaling 4.1.1 support for AMD Radeon RX 7000 Series GPUs, and FSR Frame Generation 4.0.1 improvements. ### Source excerpt The updated AMD FSR™ Unreal® Engine plugin brings ML-powered upscaling and frame generation to Unreal Engine 5.8, now extending FSR Upscaling support to AMD Radeon RX 7000 Series GPUs. ## What I Saw at ICML 2026 DevFeed: [What I Saw at ICML 2026]() Original publisher: [Read original article]() Author: zj-karina (Яндекс) Published: 2026-08-25T07:01:28Z Content type: article Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [Machine learning](), [Artificial Intelligence](), [Large Language Model]() Tags: [ai](), [icml](), [icml-2026](), [llm-agents](), [ml](), [rlhf](), [tag-316edb31b5b3](), [tag-44d9110ce940]() ### AI overview A Yandex developer reports from ICML 2026 in Seoul, describing the conference format, its scale, Yandex research presented there, and discussions about AI agents. ### Source excerpt Зачем тратить сутки на перелёты, мчаться на другой конец света и жить неделю в режиме нон-стоп на одной из главных ML-конференций планеты, когда пейпер уже на arXiv, код -- на GitHub, а краткие выжимки из выступлений -- мгновенно в соцсетях? Меня зовут Карина Романова, я разработчик в Яндексе и занимаюсь LLM-агентами в Алисе. В июле мы с командой прилетели в Сеул на ICML 2026, и я ответила себе на вопрос "зачем?". Для нас офлайн-конференции -- это единственный способ за несколько дней прочувствовать реальный фокус сообщества, встретиться с авторами работ и узнать детали, которых нет в опубликованных текстах. В этой статье расскажу, как устроена ICML изнутри, чем запомнилась программа этого года, какие наши исследования вызвали наибольший ажиотаж и почему заметная часть разговоров на конференции снова вращалась вокруг AI-агентов. Читать далее ## Backblaze Drive Stats: How an Open Dataset Powers Academic and AI/ML Research DevFeed: [Backblaze Drive Stats: How an Open Dataset Powers Academic and AI/ML Research]() Original publisher: [Read original article]() Author: Stephanie Doyle Published: 2026-08-13T15:08:38Z Content type: article Language: en Sources: [Backblaze Blog | Cloud Storage & Cloud Backup]() Topics: [dataset](), [DRIVE](), [Machine learning](), [Transformer](), [Disk image]() Tags: [ai-ml](), [articles](), [b2cloud](), [cloud-storage](), [data](), [drive](), [featured](), [featured-cloud-storage](), [hard-drive-stats](), [ml](), [research](), [source](), [storage]() ### AI overview The article explains how Backblaze Drive Stats evolved from an internal hard-drive reliability tool into an open dataset used in academic and AI/ML research. It describes the dataset's real-world scale, quarterly publication, SMART attributes, labeled failures, broad manufacturer coverage, and use in hard-drive failure prediction research. ### Source excerpt Backblaze Drive Stats has been cited in more than 105 academic papers and AI/ML projects since 2018. Explore the research it powers and download the dataset. The post Backblaze Drive Stats: How an Open Dataset Powers Academic and AI/ML Research appeared first on Backblaze Blog | Cloud Storage & Cloud Backup ## MiniDXNN v0.4.0: Interactive neural texture compression on DirectX 12 DevFeed: [MiniDXNN v0.4.0: Interactive neural texture compression on DirectX 12]() Original publisher: [Read original article]() Author: Takahiro Harada; Sho Ikeda Published: 2026-08-13T14:30:00Z Content type: release Language: en Sources: [AMD GPUOpen]() Topics: [Compression](), [GPU](), [Inference](), [mlp](), [Machine learning](), [GUI](), [shaders](), [Open Source]() Tags: [agility-sdk](), [announcement](), [compression](), [directx](), [driver](), [getting-started](), [gpu](), [gpu-open-sdks](), [gpu-open-tools](), [gpuopen-sdks](), [gpuopen-tools](), [graphics-apis](), [gui](), [inference](), [maths](), [memory](), [microsoft](), [microsoft-agility-sdk](), [microsoft-directx](), [ml](), [mlp](), [neural](), [open-source](), [product-release](), [quick-start](), [real-time](), [shaders](), [technical-article](), [technical-articles](), [tools]() ### AI overview MiniDXNN v0.4.0 is an open-source library for GPU-accelerated MLP inference and training on DirectX 12. The release adds D3D12 Linear Algebra support, input encoding for neural texture compression, and a real-time GUI application for training and visualizing texture representations. ### Source excerpt MiniDXNN v0.4.0 introduces D3D12 Linear Algebra (SM 6.10) support, input encodings and neural texture compression, plus a real-time GUI app that trains and visualizes GPU-accelerated MLPs on DirectX® 12. ## Forking-Sequences -- Part II: Multi-Horizon Forecast Ensembling with Reduced Volatility DevFeed: [Forking-Sequences -- Part II: Multi-Horizon Forecast Ensembling with Reduced Volatility]() Original publisher: [Read original article]() Author: Willa Potosnak Published: 2026-08-10T21:52:23Z Content type: article Language: en Sources: [ML@CMU]() Topics: [Sequences](), [Time Series](), [foundation-models](), [Inference](), [Machine Learning & Artificial Intelligence]() Tags: [ai](), [artificial-intelligence](), [automated-machine-learning](), [big-data](), [computer-science](), [data-science](), [deep-learning](), [ethics-and-ai](), [forecasting](), [foundation-models](), [inference](), [machine-learning](), [ml](), [multi-horizon-forecasting](), [natural-language-processing](), [neural-networks](), [paper](), [research](), [sequences](), [time-series]() ### AI overview The article describes forecast ensembling for multi-horizon time-series models. Exponential smoothing reduces forecast volatility by about 10-13% with less than 0.1% accuracy degradation, including in zero-shot use with pretrained time-series foundation models. ### Source excerpt Based on: Potosnak, W., Wolff, M., Cao, M., Ma, R., Konstantinova, T., Efimov, D., Mahoney, M.W., Oreshkin, B., & Olivares, K.G. "Forking-Sequences: Statistically and Computationally Efficient Multi-Horizon Forecasting with Reduced Volatility." Transactions on Machine Learning Research, 2026. (Disclaimer: Code implementation not used in the paper; not affiliated with Amazon -- provided as a reference for forking-sequences and forecast ensembling) TL;DR Ensembling, nearly for free. Forking-sequences already produces overlapping forecasts for every target date across FCDs in a single forward pass, so ensembling them at inference adds no extra encoder computation compared with window-sampling. Two new forecast volatility metrics. scaled Forecast Percentage Change (sFPC) measures raw revision size in real time (no ground truth needed); Excess Volatility (EV) goes further, rewarding accuracy-improving revisions and only penalizing the ones that move forecasts away from the truth or overshoot it. Reduced volatility without sacrificing accuracy. Exponential-smoothing forecast ensembling (α = 0.9) reduces sEV by 10-13% across all encoder types, with less than 0.1% accuracy degradation. Works zero-shot on models pretrained with window-sampling. Forecast ensembling applied to pretrained Time Series Foundation Models (TSFMs) -- Chronos-2, Toto 2.0, TimesFM, PatchTST, N-BEATS -- cuts volatility by ~10% with negligible accuracy cost (less than 0.1%). In Part [...] ## How Yandex combined image and document text signals for image-search ranking DevFeed: [How Yandex combined image and document text signals for image-search ranking]() Original publisher: [Read original article]() Author: nikolaevkona (Яндекс) Published: 2026-08-10T08:00:19Z Content type: tutorial Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [яндекс](), [vlm](), [Image](), [realtime]() Tags: [image](), [machine-learning](), [ml](), [realtime](), [tag-4004cf5948d3](), [tag-8be2dbf54d97](), [vlm](), [yandex-e983188bc433]() ### AI overview Yandex describes using multimodal vision-language models to jointly assess an image and its associated document text for image-search ranking. The team distilled a larger model into lighter models for different pipeline stages and reports deployment in real-time search. ### Source excerpt Исторически в Яндекс Картинках релевантность документа оценивалась по двум сигналам: насколько запросу подходит само изображение и насколько -- текст, связанный с этим изображением. Такой подход позволяет учесть контент картинки и не провалиться на визуально трудноотличимых объектах, однако он же порождает проблему: в "серой зоне", когда текстовая релевантность не сонаправлена с картиночной, становится неочевидно, как именно агрегировать сигналы в финальный скор релевантности. Привет! Я Константин Николаев, занимаюсь внедрением нейротехнологий в Поиске по картинкам. В этой статье я расскажу, как наша команда научила модели смотреть на картинку и читать текст документа одновременно: начали с тяжёлой мультимодальной VLM ради максимального качества, а затем дистиллировали её в набор лёгких моделей -- по одной под каждую стадию пайплайна. Что из этого удалось довести до realtime-поиска с десятками тысяч запросов в секунду и как совместный анализ двух модальностей добавил 5% релевантных картинок в топ выдачи -- под катом. Читать далее ## 🗓 This Week In AI Research (25-31 July 26) DevFeed: [🗓 This Week In AI Research (25-31 July 26)]() Original publisher: [Read original article]() Author: Dr. Ashish Bamania Published: 2026-08-07T01:00:43Z Content type: article Language: en Sources: [Into AI]() Topics: [releases](), [Machine learning](), [deepseek](), [foundation-models](), [Mixture of Experts (MoE)](), [lora](), [codex]() Tags: [ai](), [deep-learning](), [deepseek](), [lora](), [ml](), [moe](), [performance](), [releases](), [research]() ### AI overview A weekly roundup of AI research and releases covering DeepSeek-V4-Flash-0731, the Pangram 4 AI-text classification model, the OpenMLE system and its Frontis-MA1-35B agent, and the Metis memory foundation model. ### Source excerpt The top 10 AI research papers and releases this week. ## Чем запомнилась ICRA 2026: Reinforcement Learning, генерация сложных сценариев поведения и будущее робототехники DevFeed: [Чем запомнилась ICRA 2026: Reinforcement Learning, генерация сложных сценариев поведения и будущее робототехники]() Original publisher: [Read original article]() Author: egavolk (Яндекс) Published: 2026-08-04T08:00:45Z Content type: article Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [Reinforcement learning](), [Robotics](), [Simulation](), [Machine learning]() Tags: [2026](), [icra](), [ml](), [reinforcement-learning](), [rl](), [robotics](), [simulation](), [tag-511fbf58fd45](), [tag-6faff4be08e9](), [tag-d704a344cc75](), [tag-dace475544fb]() ### AI overview The article reviews notable trends, papers, and engineering trade-offs discussed at ICRA 2026, with emphasis on reinforcement learning, autonomous-vehicle perception and planning pipelines, simulation, rare edge-case generation, and robotic learning. It also discusses award-winning work on manipulation, humanoid robots, and camera-conditioned policy learning. ### Source excerpt Привет, Хабр! В начале июня в Вене прошла главная международная конференция по робототехнике и автономным системам -- International Conference on Robotics and Automation (ICRA). В этом году среди участников была и наша команда автономного транспорта Яндекса. Топиков, которые обсуждаются на ICRA, много, потому что она не только об ML -- она скорее о робототехнике в целом. Например, есть секции о механизмах и дизайне, а также о медицинских роботах. Было немало и чисто инженерных работ. Ключевой топик докладов на конференции -- RL, он же Reinforcement Learning, обучение с подкреплением. Также нас интересовали статьи по классическому пайплайну автономного автомобиля: perception + prediction + planner + simulation. Новые подходы к Robotic Learning тоже интересны, так как их можно перенести на задачи автономного транспорта. Меня зовут Егор Волков, я занимаюсь претрейном модели планирования движения в автономном транспорте Яндекса. Вместе со мной на конференцию ездил Максим Спорышев -- руководитель службы поведения и предсказания движения. В этой статье мы собрали самые интересные тренды, доклады и инженерные развилки, которые заметили на ICRA 2026, -- от Reinforcement Learning и генерации редких edge-кейсов до того, куда вообще двигается ML в робототехнике. Читать далее ## Using Activity isolation as a security boundary DevFeed: [Using Activity isolation as a security boundary]() Original publisher: [Read original article]() Author: Houman Kargaran Published: 2026-08-04T00:00:00Z Content type: tutorial Language: en Sources: [Temporal Blog]() Topics: [Security](), [pii](), [Architecture & Design](), [MCP Server](), [data](), [Machine learning]() Tags: [community](), [engineering](), [isolation](), [mcp](), [mcp-server](), [ml](), [pii](), [security]() ### AI overview This guest post explains why a regulated application keeps PII scanning, classification, redaction, and storage within a single Temporal Activity. The design prevents sensitive data from crossing Activity boundaries and appearing in the Temporal UI, while using an in-house MCP server, an internal ML model, and a swappable storage layer. ### Source excerpt How keeping PII scan, classification, and storage inside one Temporal Activity stops raw data from ever crossing a boundary. ## Post-mortem GPU crash debugging with LLMs DevFeed: [Post-mortem GPU crash debugging with LLMs]() Original publisher: [Read original article]() Author: Amit Ben-Moshe; Amit Mulay Published: 2026-07-14T12:00:00Z Content type: article Language: en Sources: [AMD GPUOpen]() Topics: [GPU](), [debugging](), [Model Context Protocol](), [MCP Server](), [Large Language Model](), [Code](), [Post Mortem]() Tags: [announcement](), [coding-agents](), [debugging](), [getting-started](), [gpu](), [gpu-open-tools](), [gpuopen-third-party](), [gpuopen-tools](), [graphics-apis](), [llms](), [mcp](), [mcp-server](), [memory](), [microsoft](), [ml](), [news](), [product-release](), [quick-start](), [radeon-developer-tool-suite](), [radeon-gpu-detective](), [rdts](), [rgd](), [root-cause-analysis](), [technical-article](), [technical-articles](), [third-party](), [tools](), [user-guides-manuals]() ### AI overview This article introduces the open-source AMD Radeon GPU Detective MCP Server, which gives LLMs structured access to GPU crash dumps and application source code for post-mortem debugging. It describes a workflow in which an LLM investigates crash evidence and suggests source-code fixes through a single natural-language prompt. ### Source excerpt The new AMD RGD MCP Server connects LLM agents to AMD's GPU crash analysis pipeline, turning a single prompt into root-cause analysis and source-code fix suggestions. ## DoorDash's Personalization Stack Uses Semantic Memory, Embeddings, and Context Graphs DevFeed: [DoorDash's Personalization Stack Uses Semantic Memory, Embeddings, and Context Graphs]() Original publisher: [Read original article]() Author: Alexandre Zajac Published: 2026-07-13T15:30:43Z Content type: article Language: en Sources: [Hungry Minds]() Topics: [personalization](), [Embeddings](), [Language models](), [Large Language Model](), [data](), [Graphs]() Tags: [ai](), [data](), [embeddings](), [graph](), [llms](), [ml](), [personalization](), [software]() ### AI overview The article describes DoorDash's unified memory platform for personalization. It explains how behavioral signals are converted into semantic memory using layered context, LLM-synthesized memory blocks, versioned manifests, asymmetric dense embeddings, and a consumer context graph. ### Source excerpt PLUS: Uniqlo Decoded 🚨, Agentic patterns⚡, Be the idiot mindset 👨💻 ## Join me as I go live tomorrow to discuss how to keep up with ML research DevFeed: [Join me as I go live tomorrow to discuss how to keep up with ML research]() Original publisher: [Read original article]() Author: Dr. Ashish Bamania Published: 2026-07-10T15:00:23Z Content type: article Language: en Sources: [Into AI]() Topics: [Machine learning]() Tags: [ml](), [research]() ### AI overview An announcement for a free livestream on July 11, 2026, about strategies for keeping up with machine learning research. The session will cover evaluating research quality, reading papers efficiently, and extracting practical takeaways. ### Source excerpt 👋🏻 Hey friend! ## From Traditional ML to AI Agents: How Booking.com Scales AI Observability With Arize AI DevFeed: [From Traditional ML to AI Agents: How Booking.com Scales AI Observability With Arize AI]() Original publisher: [Read original article]() Author: Amir Bitaraf Published: 2026-07-10T07:52:18Z Content type: article Language: en Sources: [Booking.com Data Science]() Topics: [ai observability](), [agent observability](), [Machine learning](), [observability](), [human review]() Tags: [agents](), [ai](), [ai-agents](), [ai-observability](), [blog-posts](), [data-quality](), [human-review](), [latency](), [ml](), [observability]() ### AI overview Booking.com describes building an AI-native observability stack for traditional machine learning systems and agentic AI workflows. The article explains that its diverse systems require observability to detect changes, regressions, data quality issues, misconfigurations, and responsible-AI concerns across different operating constraints and user contexts. ### Source excerpt Building an AI-native observability stack for agentic AI and traditional ML at Booking.com Authors: Amir Bitaraf, Shahaf Veber Why AI Observability Matters at Booking.com At Booking.com, AI helps travellers and partners in every step of their journey, from how people discover destinations to the way we support them while they're on the road. Rather than a single flagship model, we rely on a large and growing collection of systems that each solve a specific problem at scale. To make this concrete, consider a few examples: Trip planning assistants that help travelers turn vague ideas ("somewhere warm in April with good hiking") into concrete, bookable itineraries. On-site helpers that turn property details, amenities, reviews, and options into plain-language guidance, so people can choose the right stay with confidence. Partner copilots that help accommodation partners and other suppliers respond to guest messages faster and more consistently, while still staying in control of the final reply. Ranking systems that decide which options to show first in search and recommendation to surfaces, balancing user relevance with experimentation needs. Fraud detection models that quietly protect customers and partners in the background by flagging suspicious activity before it turns into real harm. Each of these systems is built and iterated on by different teams, uses different data, and runs under different constraints such as real-time vs batch, strict latency budgets vs more relaxed ones, fully automated vs human-in-the-loop. As we scale this ecosystem, observability becomes a first-class requirement, not a nice-to-have as we need to: Know when something changes in the real world, a new travel pattern, a data quality issue, a misconfiguration and how that affects model behaviour and user experience. Detect regressions early: slower responses, more confusing answers, drops in relevance or conversion, or subtle shifts that only show up for specific geographies, devices, or use ## Как оптимизировать инференс LLM: кеширование, время ответа и GPU-ресурсы DevFeed: [Как оптимизировать инференс LLM: кеширование, время ответа и GPU-ресурсы]() Original publisher: [Read original article]() Author: hotckisss (Яндекс, Yandex Cloud & Yandex Infrastructure) Published: 2026-07-08T07:04:08Z Content type: tutorial Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [Large Language Model](), [Inference Performance](), [Low-Latency Inference](), [GPU](), [Artificial Intelligence]() Tags: [ai](), [inference](), [llm](), [ml](), [mlops](), [tag-b92bf5906bbd](), [time](), [token]() ### AI overview This article explains how to optimize LLM inference in production under mixed workloads. It examines causes of increased Time to First Token, including scheduling, GPU memory allocation, streaming output queues, networking, long contexts, and KV-cache growth, and outlines optimizations such as separating prefill and decode, contextual parallelism, speculative decoding, cache-aware load balancing, and large-model delivery. ### Source excerpt Вы запустили LLM-инференс в продакшене. Поток запросов не менялся, нагрузка та же, что вчера, -- а Time to First Token внезапно вырос в три раза. Первая мысль: что-то с моделью. На деле причина почти никогда не в модели -- она прячется в планировщике, аллокаторе GPU-памяти, очереди стримингового вывода или сети. Чем длиннее контекст, тем больнее. Для классического attention вычислительная сложность растёт очень быстро. KV-кеш раздувается до десятков и сотен гигабайт, а в облаке всё это происходит на совершенно произвольном трафике: у одного клиента кодовый ассистент, у другого -- аналитика на миллион запросов в день, у третьего -- голосовой робот. На таких смешанных нагрузках всплывает то, чего не видно на референсных замерах вендора. Читать далее ## Помочь пользователю открыть новое: как мы боролись с замкнутым кругом рекомендаций в Яндекс Лавке DevFeed: [Помочь пользователю открыть новое: как мы боролись с замкнутым кругом рекомендаций в Яндекс Лавке]() Original publisher: [Read original article]() Author: ramilboiarchenkov (Яндекс) Published: 2026-07-07T07:02:47Z Content type: tutorial Language: ru Sources: [Яндекс - Как мы делаем Яндекс / Статьи]() Topics: [Machine learning]() Tags: [discovery](), [exploration](), [machine-learning](), [ml](), [tag-4004cf5948d3](), [tag-b0a411324cb6](), [tag-b2cbb9058e3c]() ### AI overview This article explains how the Yandex Lavka team addressed the feedback loop in recommendation systems, which can overemphasize familiar products and limit discovery. It describes personalized exploration of unfamiliar products and discusses calibrating the probability and aggressiveness of that exploration for each user. ### Source excerpt Хорошая рекомендательная система быстро учится угадывать, что вы положите в корзину. И чем точнее она угадывает, тем реже показывает что-то незнакомое: ведь выгоднее предлагать проверенное. Со временем система замыкается на привычках человека и перестаёт показывать ему хоть что-то за их пределами. Беда в том, что интересы меняются, а система просто так этого не замечает. Изменить ситуацию, как правило, удаётся лишь ценой краткосрочных потерь: стоит добавить в выдачу незнакомые товары, и объём ближайших покупок неизбежно начинает снижаться. Меня зовут Рамиль Боярченков, я занимаюсь машинным обучением в команде Яндекс Лавки. Расскажу, как мы собрали механизм, который подмешивает незнакомые товары персонально -- тем, кто к ним расположен, -- и с какой вероятностью это делать для каждого пользователя. По пути разберу, как мы калибровали "агрессивность" exploration и что получилось в итоге. Читать далее [Next page]()