Contact
LOG IN
Sign up
Webinar

Scale Vision-Language-Action (VLA) pipelines with Anyscale and Weights & Biases

  • May 12, 2026
  • 10am PT
  • Virtual
Register Now
Event Overview
Simulation-based evaluations accelerate robotics development by orders of magnitude, replacing hardware evaluation constraints with simulated rollouts. A production Vision-Language-Action (VLA) model fine-tuning pipeline must independently scale three core components across heterogeneous hardware: data loading on CPUs, training on one class of GPUs, and policy evaluation on another. Each sim rollout is stochastic, so meaningful evaluation requires running hundreds or thousands of simulations with domain randomization to ensure real-world transfer and generalization. Without the right tools to orchestrate these workloads and track results at scale, moving from simulation to real-world deployment lacks the control and visibility needed to deploy with confidence. This session covers how Anyscale and Weights & Biases close these gaps for physical AI teams.
What to expect

Join Omar Shorbaji, Solutions Engineering Manager, ML Platforms, Anyscale, and Anushrav Vatsa, AI Solutions Engineer, Weights & Biases, for a practical look at how to orchestrate VLA pipelines end-to-end with Ray and how to track it all with W&B Models.

What you'll learn
  • Why physical AI pipelines look fundamentally different from standard ML and what that means for your infrastructure
  • How Anyscale helps you orchestrate heterogeneous training + sim-eval pipelines, independently scaling each stage on the right hardware
  • How to make sense of stochastic sim-eval at scale, aggregating results across hundreds of rollouts and connecting them back to training decisions with W&B
  • Patterns for tracking rich sim artifacts (video rollouts, trajectory data, task success distributions) alongside training metrics
  • A reusable blueprint you can adapt to your own robotics workflow
Who should attend

AI/robotics researchers and scientists, mechanical and electrical engineers, AI engineers, software developers, and AI leaders.

Featured
speakers

Omar Shorbaji
Omar
Shorbaji
Solutions Engineering Manager, ML Platforms
Anyscale
Anu Vastra
Anushrav
Vatsa
Sr. AI Solutions Engineer
Weights & Biases