Examples# Multi-modal AI pipeline Development Production No infrastructure headaches LLM training and inference Set up Data ingestion Distributed fine-tuning Batch inference Online serving Production Audio batch inference Prerequisites Setup Streaming data ingestion Audio preprocessing GPU inference with Whisper LLM-based quality filter Persist the curated subset Distributed XGBoost pipeline Time-series forecasting Setup Acknowledgements Scalable video processing Distributed RAG pipeline Notebooks Deploy MCP servers Why Ray Serve for MCP Anyscale service benefits Prerequisites Development Production No infrastructure headaches Build a tool-using agent Architecture overview Dependencies and compute resource requirements Implementation: Building the services Deploy the services Test the agent Next steps Build a multi-agent system with the A2A protocol 1. Architecture 2. Project structure 3. Get started with local deployment 4. Deploy to production on Anyscale 5. Deep dive: Understand each component 6. Next steps 7. Additional resources