AI breakthroughs depend on more than models and chips. The compiler layer is becoming a major source of speed and efficiency gains. #mlresearchpapers #mlinfrastructure...Show more

Join @clockworkio & @linuxfoundation for a free live webinar TOMORROW at 9:00 AM PT: "Handling Hardware Failures During Training: A Comparative Analysis of Fault Tolerant Training Frameworks". Learn more & register: bit.ly/4cnpcuA #OpenSource #Linux #DistributedTraining #FaultTolerance #MLInfrastructure #MLOps

Join @clockworkio & @linuxfoundation for a free live webinar on Thursday, April 23 at 9:00 AM PT: "Handling Hardware Failures During Training: A Comparative Analysis of Fault Tolerant Training Frameworks". Learn more & register: bit.ly/4cnpcuA #OpenSource #Linux #DistributedTraining #FaultTolerance #MLInfrastructure #MLOps

How do ML teams actually share GPU clusters when demand outstrips supply? Three dominant allocation models keep appearing: 1️⃣ Quota-based allocation (virtual clusters) Give each team a guaranteed slice. Others can borrow your idle capacity, but you can reclaim it. Microsoft's HiveD takes this further with "topology-aware" guarantees so distributed training jobs get GPUs that can actually talk to each other efficiently. 2️⃣ Priority tiers with preemption Define clear priority classes (production vs batch vs dev). When contention hits, lower-priority work gets evicted. Uber's Peloton combines this with quotas: you get your share, but production can always kick out batch jobs borrowing from your pool. Catch: requires checkpoint-friendly workloads, otherwise preempted jobs waste all their compute. 3️⃣ Time-windowed access Reserve GPU access in scheduled blocks (e.g., "your 2-week slot on the training cluster"). Meta uses this for major training runs, forcing teams to arrive prepared with data staged and hypotheses tested. Most mature orgs end up with a hybrid. The interesting part isn't which algorithm wins. It's how teams build trust in the rules. #MLOps #GPUCompute #MachineLearning #MLInfrastructure

Most GPU scheduling advice assumes you have "one problem." You don't. You have a combination of two variables: Job shape: → High-frequency small jobs (sweeps, experiments, iteration) → Low-frequency mega jobs (multi-day distributed runs) Scarcity level: → Mild: queues are annoying but survivable → Severe: demand blocks roadmaps, people game the system If you're in Europe, you're probably starting in "severe" by default. The continent has <5% of new AI-optimised compute capacity vs ~70% in the US. Top-tier GPU families are available in fewer regions, often requiring account-team access just to provision. This isn't a complaint. It's a design constraint. European ML teams have to get governance right earlier because you can't just "route around" scarcity by shifting regions. The right allocation model depends on where you sit: High-frequency mild scarcity: Quotas, fair-share, basic priorities. Don't over-engineer. High-frequency severe scarcity: Preemption contracts, idle reclamation, anti-squatting policies. Otherwise people hoard defensively. Mega jobs any scarcity: Time windows, reservations, topology-aware placement. Explicit governance for "who gets the big slot." Mixed workloads: Separate lanes plus elastic borrowing with clear preemption contracts. The evidence: Microsoft's Philly cluster found 78.4% of delays for large jobs came from fragmentation, not raw shortage. Meta saw a single 1024-GPU job failure trigger 548 preemptions. GPU scheduling isn't a queuing problem. It's a governance problem. And under scarcity, compute stops behaving like elastic cloud and starts behaving like capital equipment. #GPUScheduling #MLInfrastructure #MLOps

More resilient training. Less communication overhead. ​That's the power of CheckFree (Fault Tolerance) and SkipPipe (Pipeline Efficiency). ​gensyn is tackling the hardest bottlenecks in distributed ML. ​#AI #MLInfrastructure

🚀 Yotta v0.6.0 New Features! We're making it even easier to launch, manage, and customize your AI workloads. Here's what’s new: 🌍 Multi-region selection – Deploy across multiple zones with ease 🧠 CPU memory filters – Find the right pods for memory-heavy workloads 🔒 HTTPS for Jupyter – Enjoy fully encrypted, end-to-end notebook access 🖼️ New Launch Templates – Spin up pods with ComfyUI pre-installed 🐳 Self-Hosted Docker Registry – Bring your own images, securely ⚙️ Custom Launch Specs – Save & reuse your preferred pod configs Ready to try it out? 👉 console.yottalabs.ai #YottaLabs #DeAI #GPUCloud #AIInfra #MLOps #Jupyter #ComfyUI #Docker #MLInfrastructure

How can you write distributed training jobs that fail gracefully and recover intelligently? In this example, we show how Runhouse's Kubetorch gives programmatic control over training. When the model OOMs, the error propagates back to the driver program, where you retain full control over what happens next. This allows for smart, dynamic adjustment (like reducing batch size or relaunching on larger compute) without needing to restart the whole job. You can see the code in our GitHub repo: lnkd.in/eqQy-A-8 #MLInfrastructure #PyTorch #MLOps

HIRING: Senior Technical Program Manager - AI Research Systems / US, CA, Santa Clara 💰 USD 124K 👉 aijobs.net/J473946/ #Engineering #KPIs #MachineLearning #MLinfrastructure #Modeltraining #Research

HIRING: Director, AI and ML Platform / US, CA, Santa Clara 💰 USD 308K 👉 aijobs.net/J473976/ #Engineering #MachineLearning #MLinfrastructure #Modeltraining #Research

🌟 Prasanna Ganesan, CEO at @MachinifyAI , Inc. joins me at @BatteryVentures offices for our latest podcast episode! Prasanna is backed by Battery Ventures’s @Dthakker02 , @GV (Google Ventures) and @matrixvc . They’ve achieved astonishing results given the difficulty of the healthcare market for startups: $200B Medical claims reviewed annually, 52M lives impacted by Machinify, and working with 4 of the Top-10 largest payers. 🌐 In this episode you’ll learn how AI companies can catch cost curves early to deliver products cheaper than competitors. In addition, companies will learn how to craft their go to market in order to prove the efficacy of their AI platform to large customers in high stakes environments like healthcare. WATCH HERE Youtube 👉 youtu.be/KnWa_6qrvLA Apple Podcast 👉 podcasts.apple.com/us/podcas… Spotify 👉 open.spotify.com/show/7JSs9N… Podcast Website 👉 collectiveintelligenceai.com… 🎤 @CollectPod by @GenAICollective brings the brightest minds in AI to the table to share our lessons in the trenches! Each interview provides a unique perspective on this burgeoning industry, delivering sharp insights and informed commentaries shared to our community of founders, funders, and thought leaders. CC: @chappyasel, @PKelaita , @matthuangbrain 📢 SHARE, LIKE, COMMENT #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology #software #AI

Must watch episode with Shane Orlick from @heyjasperai and The @GenAICollective 's Thomas Joshi (@gradientguide) to discuss the future of AI and automation on the Collective Intelligence AI Podcast! #artificialintelligence #machinelearning #deeplearning #aialignment #aisafety #mlops #mlinfrastructure #technology #software

🚀 Mind-expanding conversation with Shane Orlick from @heyjasperai and Thomas Joshi (@gradientguide) on our latest Collective Intelligence AI Podcast episode by The @GenAICollective ! #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology #software #AI #venturecapital

🌟 Shane Orlick, President at @heyjasperai joins me for our latest @collectpod episode! The AI platform for marketers has partnered with investors like @insightpartners, @BessemerVP , @IVP , @FoundationFund , @FoundersCC , @coatuemgmt and @HubSpot Ventures, @amasad, @ShaanVP, @collinmathilde, @ClementDelangue, and Anthony Maslowski from @cerebras . Founded by @DaveRogenmoser 🚀, Chris Hull, and JP Morgan, @heyjasperai has since served customers like @WalkMeInc , @CloudBees , @Amplitude_HQ , and @bloomreach_tm WATCH HERE Youtube 👉 youtu.be/iSW_asvexfo Apple Podcast 👉 podcasts.apple.com/us/podcas… Spotify 👉 open.spotify.com/show/7JSs9N… Podcast Website 👉 collectiveintelligenceai.com… TOPICS DISCUSSED 🎨 AI Revitalizes Creativity: Shane highlights how AI is reigniting creativity and execution in technology, making it an exciting time to innovate. 🌍 Empathy Bridges AI Adoption: His experience reveals that empathy towards those adapting to AI is crucial for successful technological integration. 🤗 Ethics Guide AI Advancement: Shane underscores the importance of ethical frameworks and governance in AI's evolution, ensuring it enhances rather than replaces human creativity. 🎤 The Collective Intelligence AI Podcast by @GenAICollective brings the brightest minds in AI to the table to share our lessons in the trenches! Each interview provides a unique perspective on this burgeoning industry, delivering sharp insights and informed commentaries shared to our community of founders, funders, and thought leaders. 📢 SHARE, LIKE, COMMENT #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology #software #AI #venturecapital

🌟 @BainCapVC, @slaterstich joins me for the latest @CollectPod podcast episode by the @GenAICollective! 🔗 Watch the Episode Here: Youtube 👉 youtu.be/cJOazecGQKE Apple Podcast 👉podcasts.apple.com/us/podcas… Spotify 👉 open.spotify.com/show/7JSs9N… Podcast Website 👉 collectiveintelligenceai.com… 🌐 As we explored the evolution from the "modern data stack" hype to a mature phase of integration, it's clear that sustainable, scalable solutions are the future. Plus, we discussed the journey towards operationalizing AI agents, highlighting the critical need for advanced machine learning models, comprehensive data integration, and ethical AI governance. 🔍 Slater and I also tackled hot topics like the implications of AI on startups, the significance of 'last mover advantage' in AI infrastructure, and the power law distribution in VC returns. This conversation is a must-listen for anyone interested in the cutting-edge of technology and its impact on our world. 🎤 The Collective Intelligence Community Podcast by The GenAI Collective brings the brightest minds in AI to the table to share our lessons in the trenches! Each interview provides a unique perspective on this burgeoning industry, delivering sharp insights and informed commentaries shared to our community founders, funders, and thought leaders. 📢 Share, like, and comment what your favorite moments were! #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology #software #AI #GenAICollective #AI

🌟 @BrandonGleklen, Principal at @BatteryVentures joins @TomJoshi2 and Stephen Campbell for our latest podcast episode! 🚀 🔗 Watch the Episode Here Apple Podcast 👉podcasts.apple.com/us/podcas… Spotify 👉 open.spotify.com/show/7JSs9N… Podcast Website 👉 collectiveintelligenceai.com… Youtube 👉 youtu.be/5BcuWebfwK0 Join us as we explore the intricate world of AI, dissecting key topics that are shaping the future of technology and investment. Here's a sneak peek: 🛡️ AI Moats: Necessity or Overemphasis? We dive deep into the importance of 'moats' in the AI sector. Unlike traditional industries, AI demands unique defenses like proprietary algorithms and exclusive data to stay ahead. Are people too focused on competitive moats? Brandon sheds light on this critical debate. 🕰️ Beyond the "Horseless Carriage" Era in AI Design: Are current AI products limited by outdated thinking? We discuss how breaking free from the 'horseless carriage' mindset can lead to revolutionary AI applications, moving from incremental improvements to radical innovations. 🤝 Copilot vs. Full Automation: In a world where AI's role varies from assisting humans to full automation, we explore the optimal approach for different sectors. Discover how copilot AI is transforming creative industries and why full automation is a game-changer. 🔍 Niche vs. Horizontal Opportunities: The AI industry stands at a crossroads between focusing on specialized, niche use cases and broad, horizontal opportunities. Brandon provides insights into how companies can navigate this decision, balancing innovation with practical applications. 🩺 Deep Dive into Healthcare AI: With a special focus on healthcare, we examine the transformative impact of AI on diagnosis, treatment, and the overall healthcare ecosystem. From enhancing precision to integrating with EHR systems, learn about the challenges and breakthroughs in this vital field. 🎤 The Collective Intelligence Community Podcast by @GenAICollective brings the brightest minds in AI to the table to share our lessons in the trenches! Each interview provides a unique perspective on this burgeoning industry, delivering sharp insights and informed commentaries shared to our community founders, funders, and thought leaders. 📢 Share, like, and comment what your favorite moments were! #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology #software #AI #GenAICollective

🌟 @twelve_labs (backed by @IndexVentures's @shardul_shah , @Techstars , NVentures, @intelcapital's @bharadwaj_avi, and @SamsungNext ) CTO Aidan Lee joins @TomJoshi2 and Stephen Campbell, co-founder of @getrevamp_ai , for a deep dive into the world of AI in video understanding! 🌐 Journey from Military Squad to AI Innovators: Discover the incredible story of Aidan and his co-founders, former squad mates in the Korean Military, as they forge Twelve Labs amidst a global pandemic, challenging the giants like @GoogleAI and @amazon . 🔍 Behind-the-Scenes of Twelve Labs' Success: Hear about Aidan's audacious plan to win the @IEEEorg ICCV VALUE Challenge, a pivotal moment that propelled their startup to the forefront of video AI technology. 🌍 AI Revolution in Video Understanding: Dive into how Twelve Labs is redefining the future of video AI, from winning prestigious competitions to developing cutting-edge technologies. 💡 A Glimpse into AI's Future: Aidan shares his vision on the next 5-10 years in AI video analysis and understanding, offering invaluable advice for aspiring AI professionals. 🎧 Tune in to unravel Aidan's unique insights, from his journey into AI post-military service to leading Twelve Labs' ambitious vision. Whether you're an AI enthusiast or a tech-savvy professional, this episode is a must-listen! 🎤 The Collective Intelligence Community Podcast by @GenAICollective brings the brightest minds in AI to the table to share our lessons in the trenches! Each interview provides a unique perspective on this burgeoning industry, delivering sharp insights and informed commentaries shared to our community founders, funders, and thought leaders. 🔗 Watch the Episode Here: Apple Podcast 👉podcasts.apple.com/us/podcas… Spotify 👉 open.spotify.com/show/7JSs9N… Podcast Website 👉 collectiveintelligenceai.com… Youtube 👉 youtu.be/MOX0caCbeag 📢 Share, like, and comment what your favorite moments were! #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology #software #AI #venturecapital #GenAICollective #AI #business

🌟 @roeschinc, CTO of @OctoAICloud (formerly @octoml ), joins @TomJoshi2 and @PKelaita for our latest podcast episode! 🚀 🔗 Watch the Episode Here: Youtube 👉 youtu.be/0u0WPrbIY80 Apple Podcast 👉podcasts.apple.com/us/podcas… Spotify 👉 open.spotify.com/show/7JSs9N… Podcast Website 👉 collectiveintelligenceai.com… 🌐 In this not-to-be-missed episode, we delve into: 🛠️ The Genesis of OctoAI: Uncover the story behind OctoAI's inception and how systems research is supercharging deep learning. 🔍 A Multidisciplinary Approach to AI: Jared shares how his diverse experience in computer science influences the development of groundbreaking ML systems, especially within Apache TVM. 🌍 Seattle's AI Scene: Dive into Jared's insights on Seattle as a burgeoning hub for AI innovation, talent, and entrepreneurship. 🤝 Open Source and Machine Learning: Explore the symbiotic relationship between open-source software and commercial AI ventures, and their collective impact on advancing ML. 👥 Building an AI Powerhouse: Learn about the qualities Jared values in team members as OctoAI paves the way in machine learning infrastructure. 💡 Innovations at OctoAI: Discover how OctoAI's recent launches are revolutionizing generative AI, and the journey since their transformative shift from OctoML. 🚀 Insider's View on AI's Future: Jared offers his unique perspective on the future of AI infrastructure, transcending beyond efficiency metrics. 🎙️ Get ready for a deep dive into the mind of a leader who's at the forefront of the AI revolution. Whether you're an AI enthusiast, professional, or just curious about the future of technology, this episode is for you! 🎤 The Collective Intelligence Community Podcast by @GenAICollective brings the brightest minds in AI to the table to share our lessons in the trenches! Each interview provides a unique perspective on this burgeoning industry, delivering sharp insights and informed commentaries shared to our community founders, funders, and thought leaders. 📢 Share, like, and comment what your favorite moments were! #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology #software #AI #venturecapital #GenAICollective #AI #business

🥓 Tune into my convo with @AdamHaney, VP of Engineering at @invtechinc and Ex-@Meta/@facebook on AI-powered services! 🤖 📢 Share, like, and comment what your favorite moments were! #artificialintelligence #machinelearning #deeplearning #mlops #mlinfrastructure #technology

What did i do today despite worsed rest and work leave? Task: convert Custom Finetuned Language model from Pytorch to Tflite for mobile engineers to improve inference response speed. What did I do? 👇 #MachineLearning #MLInfrastructure #AIEngineering #modeldeployment