Uploaded September 2026 | Updated November 2026, 3 hours ago
Miles is an enterprise-facing agentic RL framework for large-scale PyTorch post-training.
At Ray Summit 2026, Mao Cheng of RadixArk and Xinyu Zhang of Anyscale cover its end-to-end stack for SFT, LoRA SFT/RL, and agentic RL, scalable multi-turn tool-use rollouts, and integrations with PyTorch, Megatron, and Ray, plus lessons from production workloads: low-precision training, MoE RL stability, speculative rollout acceleration, and extensible hooks for custom models, rewards, environments, and deployment pipelines.
You'll leave with a clear picture of what enterprise agentic RL takes in production.
Liked this video? Check out other Ray Summit breakout session recordings: youtube.com/playlist?list=PLZNaJyYZRPdo
Subscribe to our YouTube channel to stay up-to-date on the future of AI: youtube.com/c/anyscale
🔗 Connect with us:
LinkedIn: linkedin.com/company/joinanyscale
X: https://x.com/anyscalecompute
Website: anyscale.com
Miles is an enterprise-facing agentic RL framework for large-scale PyTorch post-training.
At Ray Summit 2026, Mao Cheng of RadixArk and Xinyu Zhang of Anyscale cover its end-to-end stack for SFT, LoRA SFT/RL, and agentic RL, scalable multi-turn tool-use rollouts, and integrations with PyTorch, Megatron, and Ray, plus lessons from production workloads: low-precision training, MoE RL stability, speculative rollout acceleration, and extensible hooks for custom models, rewards, environments, and deployment pipelines.
You'll leave with a clear picture of what enterprise agentic RL takes in production.
Liked this video? Check out other Ray Summit breakout session recordings: youtube.com/playlist?list=PLZNaJyYZRPdo
Subscribe to our YouTube channel to stay up-to-date on the future of AI: youtube.com/c/anyscale
🔗 Connect with us:
LinkedIn: linkedin.com/company/joinanyscale
X: https://x.com/anyscalecompute
Website: anyscale.com










