Episodios

  • Fixing GPU Starvation in Large-Scale Distributed Training

    Fixing GPU Starvation in Large-Scale Distributed Training
    Apr 3 2026

    Kashish Mittal is a Staff Software Engineer at Uber, working on large-scale distributed systems and core backend infrastructure.Fixing GPU Starvation in Large-Scale Distributed Training // MLOps Podcast #367 with Kashish Mittal, Staff Software Engineer at Uber Join the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletterMLOps GPU Guide: https://go.mlops.community/gpuguide// Abstract Kashish zooms out to discuss a universal industry pattern: how infrastructure—specifically data loading—is almost always the hidden constraint for ML scaling.The conversation dives deep into a recent architectural war story. Kashish walks through the full-stack profiling and detective work required to solve a massive GPU starvation bottleneck. By redesigning the Petastorm caching layer to bypass CPU transformation walls and uncovering hidden distributed race conditions, his team boosted GPU utilization to 60%+ and cut training time by 80%. Kashish also shares his philosophy on the fundamental trade-offs between latency and efficiency in GPU serving.// BioKashish Mittal is a Staff Software Engineer at Uber, where he architects the hyperscale machine learning infrastructure that powers Uber’s core mobility and delivery marketplaces. Prior to Uber, Kashish spent nearly a decade at Google building highly scalable, low-latency distributed ML systems for flagship products, including YouTube Ads and Core Search Ranking. His engineering expertise lies at the intersection of distributed systems and AI—specifically focusing on large-scale data processing, eliminating critical I/O bottlenecks, and maximizing GPU efficiency for petabyte-scale training pipelines. When he isn't hunting down distributed race conditions, he is a passionate advocate for open-source architecture and building reproducible, high-throughput ML systems.// Related LinksWebsite: https://www.uber.com/~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]Connect with Demetrios on LinkedIn: /dpbrinkmConnect with Kashish on LinkedIn: /kashishmittal/
    Más Menos
    53 m
  • Spec Driven Development, Workflows, and the Recent Coding Agent Conference

    Spec Driven Development, Workflows, and the Recent Coding Agent Conference
    Mar 31 2026

    Jens Bodal is a Senior Software Engineer II working independently, focusing on backend systems, software architecture, and building scalable solutions across client projects.This One Shift Makes Developers Obsolete // MLOps Podcast #366 with Jens Bodal, Senior Software Engineer II, Independent Join the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletterMLOps GPU Guide: https://go.mlops.community/gpuguide// Abstract AI agents are shifting the role of developers from writing code to defining intent. This conversation explores why specs are becoming more important than implementation, what breaks in real-world systems, and how engineering teams need to rethink workflows in an agent-driven world.// BioJens Bodal is a senior software engineer based in Edmonds, Washington, with nine years of experience building developer tooling, internal platforms, and web infrastructure. He spent seven years as an SDE II at Amazon, working on teams including Amazon Games Studio and the AWS Events Management Platform. His work has focused on developer tooling, CI/CD systems, testing infrastructure, and improving the developer experience for teams operating production services. He is particularly interested in developer experience and the growing ecosystem of local tools that help engineers build and run AI systems on infrastructure they control.// Related LinksWebsite: https://bodal.devhttps://github.com/jensbodalhttps://www.youtube.com/watch?v=Yp7LYdbOuwE~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]Connect with Demetrios on LinkedIn: /dpbrinkmConnect with Jens on LinkedIn: /jensbodal
    Más Menos
    59 m
  • Operationalizing AI Agents: From Experimentation to Production // Databricks Roundtable

    Operationalizing AI Agents: From Experimentation to Production // Databricks Roundtable
    Mar 30 2026
    Databricks Roundtable episode: Operationalizing AI Agents: From Experimentation to Production. Join the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletterMLOps GPU Guide: https://go.mlops.community/gpuguideBig shout-out to Databricks for the collaboration!// AbstractThis panel discusses the real-world challenges of deploying AI agents at scale. The conversation explores technical and operational barriers that slow production adoption, including reliability, cost, governance, and security.The panelists also examine how LLMOps, AIOps, and AgentOps differ from traditional MLOps, and why new approaches are required for generative and agent-based systems. Finally, experts define success criteria for GenAI frameworks, with a focus on robust evaluation, observability, and continuous monitoring across development and staging environments.// BioSamraj MoorjaniSamraj is a software engineer working on the Agent Quality team. Previously, Samraj worked at Meta on ads/product classification research and AppLovin on MLOps. Samraj graduated with a BS+MS in Computer Science from UIUC, advised by Professor Hari Sundaram, where he worked on controllable natural language generation to produce appealing, interpretable science to combat the spread of misinformation. He also worked with Professor Wen-mei Hwu on accelerating LLM inference through extreme sparsification.Apurva MisraApurva is an AI Consultant at Sentick, focusing on assisting startups with their AI strategy and building solutions. She leverages her extensive experience in machine learning and a Master's degree from the University of Waterloo, where her research bridged driving and machine learning, to offer valuable insights. Apurva's keen interest in the startup world fuels her passion for helping emerging companies incorporate AI effectively. In her free time, she is learning Spanish, and she also enjoys exploring hidden gem eateries, always eager to hear about new favourite spots!Ben EpsteinBen was the machine learning lead for Splice Machine, leading the development of their MLOps platform and Feature Store. He is now the Co-founder and CTO at GrottoAI, focused on supercharging multifamily teams and reducing vacancy loss with AI-powered guidance for leasing and renewals. Ben also works as an adjunct professor at Washington University in St. Louis, teaching concepts in cloud computing and big data analytics.Hosted by Adam Becker// Related LinksWebsite: https://www.databricks.com/https://mlflow.org/~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]Connect with Demetrios on LinkedIn: /dpbrinkmConnect with Samraj on LinkedIn: /samrajmoorjani/Connect with Apurva on LinkedIn: /apurva-misra/Connect with Ben on LinkedIn: /ben-epstein/Connect with Adam on LinkedIn: /adamissimo/
    Más Menos
    1 h y 1 m
  • arrowspace: Vector Spaces and Graph Wiring

    arrowspace: Vector Spaces and Graph Wiring
    Mar 27 2026

    Lorenzo Moriondo is a Technical Lead for AI at tuned.org.uk, working on AI agent protocols, graph-based search, and production-grade LLM systems.


    arrowspace: Vector Spaces and Graph Wiring // MLOps Podcast #365 with Lorenzo Moriondo, AI Research and Product Engineer


    Join the Community: https://go.mlops.community/YTJoinIn

    Get the newsletter: https://go.mlops.community/YTNewsletter

    MLOps GPU Guide: https://go.mlops.community/gpuguide


    // Abstract

    Meet arrowspace — an open-source library for curating and understanding LLM datasets across the entire lifecycle, from pre-training to inference. Instead of treating embeddings as static vectors, arrowspace turns them into graphs (“graph wiring”) so you can explore structure, not just similarity. That unlocks smarter RAG search (beyond basic semantic matching), dataset fingerprinting, and deeper insights into how different datasets behave.


    You can compare datasets, predict how changes will affect performance, detect drift early, and even safely mix data sources while measuring outcomes.


    In short: arrowspace helps you see your data — and make better decisions because of it.


    // Bio

    With over a decade of experience in software and data engineering across startups and early-stage projects, Lorenzo has recently turned his focus to the AI-assisted movement to automate software and data operations. He has contributed to and founded projects within various open-source communities, including work with Summer of Code, where he focused on the Semantic Web and REST APIs.A strong enthusiast of Python and Rust, he develops tools centered around LLMs and agentic systems. He is a maintainer of the SmartCore ML library, as well as the creator of Arrowspace and the Topological Transformer.


    // Related Links

    Website: https://www.tuned.org.uk


    ~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

    Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

    Join our Slack community [https://go.mlops.community/slack]

    Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

    Sign up for the next meetup: [https://go.mlops.community/register]

    MLOps Swag/Merch: [https://shop.mlops.community/]


    Connect with Demetrios on LinkedIn: /dpbrinkm

    Connect with Chris on LinkedIn: /lorenzomoriondo


    Timestamps:

    [00:00] Graph Wiring for ML

    [00:32] RAG and Vector Similarity

    [08:58] Geometric Search Trade-offs

    [13:12] Vector DB Algorithm Integration

    [21:32] Feature-Based Retrieval Shift

    [26:04] Epiplexity and Embeddings

    [31:26] Epiplexity and Embedding Structure

    [40:15] Training vs Post-hoc Models

    [47:16] Discovery-Driven Development

    [51:22] Updating Mental Models

    [53:00] Vector Search vs Agents

    [55:30] Wrap up
    Más Menos
    56 m
  • Agentic Marketplace

    Agentic Marketplace
    Mar 20 2026

    Donné Stevenson is a Machine Learning Engineer at Prosus, working on scalable ML infrastructure and productionizing GenAI systems across portfolio companies.Pedro Chaves is a Data Science Manager at OLX Group, working on GenAI-powered search, personalization, and large-scale marketplace recommendations.Join the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletterMLOps GPU Guide: https://go.mlops.community/gpuguide// AbstractMarketplaces are about to get weird.

    With Pedro Chaves and Donné Stevenson: agents picking your house, negotiating deals, even talking to other agents for you.

    Less browsing. Less choice. More automation.

    Convenience… or giving up control?// BioDonné StevensonFocused on building AI-powered products that give companies the tools and expertise needed to harness the power of AI in their respective fields.

    Pedro ChavesPedro is a Data Science Manager at OLX Group, where he leads teams building machine learning solutions to improve marketplace performance, pricing, and user experience at scale.// Related LinksWebsite: https://www.prosus.com/Website: https://www.olxgroup.com/~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]MLOps GPU Guide: https://go.mlops.community/gpuguide


    Timestamps:[00:00] OLX: Disrupting Buyer-Seller Experiences[03:33] Redefining the Home-Buying Experience[07:40] User Feedback and Iterative Rollouts[11:25] Beyond Chat: Redefining Agent Use[14:03] User Trust and Education Challenges[16:47] Learning Curve for Automoto[20:05] Interactive Decision-Making with AI[24:47] Agents Simplify Buyer-Seller Search[28:14] Garage Sale Treasure Hunting[33:43] Agent Discovery Layer Needed[34:53] Agents Relying on Agents[39:48] Reducing Friction in Selling Stuff[41:39] Extracting Buyer Intent Systematically[44:49] Optimizing Delivery with Lockers[50:10] Generative AI Commerce Strategies[51:03] Improving Chat Interaction Layer
    Más Menos
    51 m
  • Durable Execution and Modern Distributed Systems

    Durable Execution and Modern Distributed Systems
    Mar 17 2026
    Johann Schleier-Smith is the Technical Lead for AI at Temporal Technologies, working on reliable infrastructure for production AI systems and long-running agent workflows. Durable Execution and Modern Distributed Systems, Johann Schleier-Smith // MLOps Podcast #364Join the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletterMLOps Merch: https://shop.mlops.community/Big shoutout to ⁨ @Temporalio for the support, and to @trychroma for hosting us in their recording studio// AbstractA new paradigm is emerging for building applications that process large volumes of data, run for long periods of time, and interact with their environment. It’s called Durable Execution and is replacing traditional data pipelines with a more flexible approach. Durable Execution makes regular code reliable and scalable.In the past, reliability and scalability have come from restricted programming models, like SQL or MapReduce, but with Durable Execution, this is no longer the case. We can now see data pipelines that include document processing workflows, deep research with LLMs, and other complex and LLM-driven agentic patterns expressed at scale with regular Python programs.In this session, we describe Durable Execution and explain how it fits in with agents and LLMs to enable a new class of machine learning applications.// Related Linkshttps://t.mp/hello?utm_source=podcast&utm_medium=sponsorship&utm_campaign=podcast-2026-03-13-mlops&utm_content=mlops-johannhttps://t.mp/vibe?utm_source=podcast&utm_medium=sponsorship&utm_campaign=podcast-2026-03-13-mlops&utm_content=mlops-johannhttps://t.mp/career?utm_source=podcast&utm_medium=sponsorship&utm_campaign=podcast-2026-03-13-mlops&utm_content=mlops-johann ~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]Connect with Demetrios on LinkedIn: /dpbrinkmConnect with Johann on LinkedIn: /jssmith/
    Más Menos
    1 h y 1 m
  • Performance Optimization and Software/Hardware Co-design across PyTorch, CUDA, and NVIDIA GPUs

    Performance Optimization and Software/Hardware Co-design across PyTorch, CUDA, and NVIDIA GPUs
    Feb 24 2026
    March 3rd, Computer History Museum CODING AGENTS CONFERENCE, come join us while there are still tickets left.https://luma.com/codingagentsChris Fregly is currently focused on building and scaling high-performance AI systems, writing and teaching about AI infrastructure, helping organizations adopt generative AI and performance engineering principles on AWS, and fostering large developer communities around these topics.Performance Optimization and Software/Hardware Co-design across PyTorch, CUDA, and NVIDIA GPUs // MLOps Podcast #363 with Chris Fregly, Founder, AI Performance Engineer, and InvestorJoin the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletterMLOps GPU Guide: https://go.mlops.community/gpuguide// AbstractIn today’s era of massive generative models, it's important to understand the full scope of AI systems' performance engineering. This talk discusses the new O'Reilly book, AI Systems Performance Engineering, and the accompanying GitHub repo (https://github.com/cfregly/ai-performance-engineering). This talk provides engineers, researchers, and developers with a set of actionable optimization strategies. You'll learn techniques to co-design and co-optimize hardware, software, and algorithms to build resilient, scalable, and cost-effective AI systems for both training and inference. // BioChris Fregly is an AI performance engineer and startup founder with experience at AWS, Databricks, and Netflix. He's the author of three (3) O'Reilly books, including Data Science on AWS (2021), Generative AI on AWS (2023), and AI Systems Performance Engineering (2025). He also runs the global AI Performance Engineering meetup and speaks at many AI-related conferences, including Nvidia GTC, ODSC, Big Data London, and more.// Related LinksAI Systems Performance Engineering: Optimizing Model Training and Inference Workloads with GPUs, CUDA, and PyTorch 1st Edition by Chris Fregly: https://www.amazon.com/Systems-Performance-Engineering-Optimizing-Algorithms/dp/B0F47689K8/Coding Agents Conference: https://luma.com/codingagents~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExploreJoin our Slack community [https://go.mlops.community/slack]Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)] Sign up for the next meetup: [https://go.mlops.community/register]MLOps Swag/Merch: [https://shop.mlops.community/]Connect with Demetrios on LinkedIn: /dpbrinkmConnect with Chris on LinkedIn: /cfreglyTimestamps:[00:00] SageMaker HyperPod Resilience[00:27] Book Creation and Software Engineering[04:57] Software Engineers and Maintenance[11:49] AI Systems Performance Engineering[22:03] Cognitive Biases and Optimization / "Mechanical Sympathy"[29:36] GPU Rack-Scale Architecture[33:58] Data Center Reliability Issues[43:52] AI Compute Platforms[49:05] Hardware vs Ecosystem Choice[1:00:05] Claude vs Codex vs Gemini[1:14:53] Kernel Budget Allocation[1:18:49] Steerable Reasoning Challenges[1:24:18] Data Chain Value Awareness
    Más Menos
    1 h y 26 m
  • Serving LLMs in Production: Performance, Cost & Scale // CAST AI Roundtable

    Serving LLMs in Production: Performance, Cost & Scale // CAST AI Roundtable
    Feb 19 2026

    Roundtable CAST AI episode: Serving LLMs in Production: Performance, Cost & Scale.


    Join the Community:

    https://go.mlops.community/YTJoinIn

    Get the newsletter: https://go.mlops.community/YTNewsletter

    MLOps GPU Guide:

    https://go.mlops.community/gpuguide


    // Abstract

    Experimenting with LLMs is easy. Running them reliably and cost-effectively in production is where things break.

    Most AI teams never make it past demos and proofs of concept. A smaller group is pushing real workloads to production—and running into very real challenges around infrastructure efficiency, runaway cloud costs, and reliability at scale.

    This session is for engineers and platform teams moving beyond experimentation and building AI systems that actually hold up in production.


    // Bio

    Ioana Apetrei

    Ioana is a Senior Product Manager at CAST AI, leading the AI Enabler product, an AI Gateway platform for cost-effective LLM infrastructure deployment. She brings 12 years of experience building B2C and B2B products reaching over 10 million users. Outside of work, she enjoys assembling puzzles and LEGOs and watching motorsports.


    Igor Šušić

    Igor is a founding Machine Learning Engineer at CAST AI’s AI Enabler, where he focuses on optimizing inference and training at scale. With a strong background in Natural Language Processing (NLP) and Recommender Systems, Igor has been tackling the challenges of large-scale model optimization long before transformers became mainstream. Prior to CAST AI, he worked at industry leaders like Bloomreach and Infobip, where he contributed to the development and deployment of large-scale AI and personalization systems from the early days of the field.


    // Related Links

    Website: https://cast.ai/


    ~~~~~~~~ ✌️Connect With Us ✌️ ~~~~~~~

    Catch all episodes, blogs, newsletters, and more: https://go.mlops.community/TYExplore

    Join our Slack community [https://go.mlops.community/slack]

    Follow us on X/Twitter [@mlopscommunity](https://x.com/mlopscommunity) or [LinkedIn](https://go.mlops.community/linkedin)]

    Sign up for the next meetup: [https://go.mlops.community/register]

    MLOps Swag/Merch: [https://shop.mlops.community/]


    Connect with Demetrios on LinkedIn: /dpbrinkm

    Connect with Ioana on LinkedIn: /ioanaapetrei/

    Connect with Igor on LinkedIn: /igor-%C5%A1u%C5%A1i%C4%87/
    Más Menos
    1 h y 6 m