Anyscale Courses

Ray Data and vLLM for Scalable Image Captioning

Discover how Ray Data and vLLM transform complex vision-language model pipelines into scalable, production-ready systems for high-throughput image captioning workflows.

• YouTube
• 31 minutes
• Self-Paced
• Free Video

CoServe - Max Performance, Minimal Compute

Discover how Cohere builds scalable AI infrastructure using vLLM optimizations, quantization techniques, and kernel engineering to achieve low-latency, high-throughput inference at minimal cost.

• YouTube
• 27 minutes
• Self-Paced
• Free Video

Inside NVIDIA Dynamo - Faster, Scalable AI Deployment

Discover how NVIDIA Dynamo revolutionizes large-scale LLM inference with smart scheduling, hierarchical memory management, and low-latency KV-cache transfer for maximum efficiency.

• YouTube
• 34 minutes
• Self-Paced
• Free Video

Inside Adobe Firefly - JIT-Embedding with Ray Serve for Faster GenAI Training

Discover how Adobe accelerated Firefly GenAI training using JIT-Embedding with Ray Serve, decoupling embedding computation from model training for better scalability and cost efficiency.

• YouTube
• 31 minutes
• Self-Paced
• Free Video

How AWS Scales Reinforcement Learning Across Thousands of GPUs

Discover how to build fault-tolerant, scalable RL systems for LLM alignment using Ray's elasticity with Amazon SageMaker HyperPod's resilience across thousands of GPUs.

• YouTube
• 32 minutes
• Self-Paced
• Free Video

Scaling Ray on Kubernetes - Pragmatic Strategies for Every Team

Master KubeRay deployment strategies to build stable, cost-efficient AI platforms on Kubernetes, solving resource chaos and enabling seamless multi-tenant ML operations.

• YouTube
• 30 minutes
• Self-Paced
• Free Video

Apple's Approach to Scalable Machine Learning Infrastructure on Ray

How Apple unifies training, inference, and data processing in a fault-tolerant, low-latency Ray framework for massive ML workloads.

• YouTube
• 24 minutes
• Self-Paced
• Free Video

Scaling Machine Learning at Tripadvisor - Our Journey with Ray and Anyscale

Discover how Tripadvisor integrated Anyscale and Ray to scale ML workloads, improve efficiency, and reduce costs across recommendation systems, personalization, and content understanding.

• YouTube
• 29 minutes
• Self-Paced
• Free Video

Ray Serve - Advancing Scalability and Flexibility

Discover Ray Serve's latest advancements for scalable AI inference: multi-model serving, enhanced performance optimizations, expanded APIs for complex pipelines, and multi-cloud deployment capabilities.

• YouTube
• 29 minutes
• Self-Paced
• Free Video

SGLang - An Efficient Open-Source Framework for Large-Scale LLM Serving

Discover SGLang's high-performance architecture for large-scale LLM serving, featuring optimized KVCache handling, scheduling strategies, and real-world production insights.

• YouTube
• 28 minutes
• Self-Paced
• Free Video

SkyRL tx - A Unified Training and Inference Engine

Discover SkyRL tx, an open-source system unifying transformer training and inference through a single REST interface, enabling cost-efficient multi-tenancy with LoRA adapters.

• YouTube
• 38 minutes
• Self-Paced
• Free Video

Ray Direct Transport - RDMA Support in Ray Core

Discover how Ray Direct Transport eliminates GPU bottlenecks by enabling zero-copy, direct device-to-device data transfer, dramatically accelerating distributed AI workloads.

• YouTube
• 32 minutes
• Self-Paced
• Free Video

Hybrid RL + Imitation Learning for Robotics with Ray

Discover how to build unified robotics ML infrastructure using Ray to seamlessly orchestrate RL and imitation learning across heterogeneous GPU clusters for real-time deployment.

• YouTube
• 33 minutes
• Self-Paced
• Free Video

Hybrid RL + Imitation Learning for Robotics with Ray at RAI Institute

Discover how Ray Train eliminates GPU bottlenecks in deep learning through async checkpointing, validation, and high-performance data ingestion for maximum throughput.

• YouTube
• 32 minutes
• Self-Paced
• Free Video

Introducing Terminal-Bench - Evaluating LLM Agents in Realistic Terminal Settings

Discover Terminal-Bench, a challenging benchmark for evaluating LLM agents in real-world terminal environments, addressing gaps in current agent evaluation methods.

• YouTube
• 31 minutes
• Self-Paced
• Free Video