As more teams scale open source AI on secure, governed infrastructure, Ray has become critical across the full AI lifecycle, from data curation to production serving and reinforcement learning. Ahead of PyTorch Conference North America 2026 in San Jose, we put together a guide highlighting key Ray-focused sessions on the schedule. Featured speakers include: Anyscale /@databricks / @UCBerkeley : Ion Stoica @anyscalecompute : Eric Tang, Sumanth Hedge, Josh Lee, Mengjin Yan @Google : Ankita Luthra, Trinadh Kotturu, Jago Macleod @LinkedIn : Tao Huang, Tommy Li @Pinterest : Gaurav Arora, Shunyao Li, Eric Wang @Uber : Ke Chen, Peng Zhang, Xandra Zhu University of Minnesota (@UMNews) : Arun Sharma Explore engineering takeaways from these technical sessions where you will learn about unified AI orchestration with Kubernetes, elastic training stacks, and scalable RL. Read the full guide and map out your schedule: Register for PyTorch Conference North America:

Thank you to everyone who showed up, spoke, and stayed throughout the vLLM Conference. Lines out the door for talks say a lot about this community. Thank you to @inferact for hosting! Session recordings are in the thread.

That's a wrap on the @vllm_project Conference at Ray Summit. Lines out the door for talks about an inference engine—that says everything about how much this community shows up. Thank you to @anyscalecompute for co-hosting with us, and we're already looking forward to the next event.

In his recent talk at @anyscalecompute's Ray Summit, Bedrock’s Vincent Gonguet covered how the company was able to get to commercial deployments of fully autonomous excavators in only two years, and ways they use Ray to power the stack: heterogeneous CPU and GPU pipelines, mining ambiguous multimodal field data, and a closed-loop simulator that models everything down to the dirt.

I watched a bunch of the talks at Ray Summit live and this one by Priunsh Syen and Tyler Titsworth at @LilaSciences about the details of their AI research platform was extremely impressive.

That's a wrap on Ray Summit 2026. Three days, 1800+ attendees, 90+ sessions, Training Day, and the first vLLM Conference. Thank you to everyone who built it with us. Keynote replays are live: Day 1: Day 2:

We stress tested Ray-2.58.0 for improvements on fault tolerance and reliability by scaling video data processing pipeline to caption 0.6 Petabytes of video data on 1600 Nvidia Blackwell GPUs in collaboration with @CoreWeave With the new release Ray Data shows: • 𝟲𝟬％ higher throughput for streaming data processing • 𝟮𝟯％ faster for batch inference • 𝟮𝟰％ faster for Ray Data shuffle This will significantly improve how world model and physical AI companies handle pre-training data. Blog:

Part 3 of the Day 2 Ray Summit keynotes - Vincent Gonguet from @bedrockrobotics is focused on advanced autonomy for physically building out the world. @anyscalecompute @raydistributed

Excited to share our Physical AI reasoning model for item stacking, CARGO! CARGO is trained with sim-to-real reinforcement learning to plan high-level actions, such as which item to pick next and where to place it, in order to maximize packing density on real-world robots. CARGO is already deployed to robots in customer operations, building stacks at over 72.5% density at over 340 sorts per hour. Read more in our technical blog: #HandleMore#