Ray Workshop: Boston

Overview of Ray's core concepts: How Ray provides simple, unified APIs for cluster computing. Ray Libraries. Setup and basic execution examples.

Learn Ray’s core concepts, including tasks, actors, and clusters. Understand how Ray scales Python and ML workloads, manage resources, and run distributed programs reliably across nodes.

Overview of data preprocessing and distributed training libraries, Ray Data & Ray Train, with an emphasis on vision language action models.

Ingest, transform, and preprocess large multimodal datasets, then create streaming pipelines to chunk data, generate embeddings at scale.

Using Ray Core to parallelize computationally intensive simulations (e.g., Mujoco, Isaac Sim)

Scale model training with data and model parallelism (e.g., FSDP). Run distributed jobs, manage checkpoints, and integrate frameworks like PyTorch with reliability and performance.