This recorded session provides a comprehensive introduction to foundation models and practical techniques for building Large Language Model (LLM) based solutions and agents. It begins with core Generative AI concepts, covering different model types and common downstream tasks such as text classification, summarization, question answering, and entity recognition.Session Topics:Foundation Models:Generative AI Concepts - ModelsUse-Cases/Downstream Tasks (eg. Text Classification, Question Answering, Summarization, Entity Recognition)Practical Techniques for building LLM and LLM AgentsPrompt engineering techniques (Zero-shot, Few-shot, CoT, ReAct with examples)Retrieval augmented generation (RAG)Model grounding techniques - contextual grounding, hallucination detection, model evaluation (LLM-as-judgeetc) Agents and components - tool calling, short-term/long-term memory, autonomous reasoningOverview of Bedrock and Lang chain components (Chains, Document Loaders, Retrievers) as building blocks for agentsTooling calling – MCP & difference from Function Calling (examples)Model Context Protocol (MCP Client & MCP Server) - Enterprise system integration & Knowledge base integrationMulti-agent collaboration – Bedrock supervisor agent and collaborator agents – A2A (Strands SDK) – modelling real-word workflows as agents – HITL systemsUse Case Discussion & DemoUse-case/case studies - GenAI use-cases in Lending, Underwriting, Payments etc. Internal use-cases for Operations etc.Demo - Financial PDF document data extraction (Bedrock Data Automation), Document creation for customer due diligence reporting and regulatory reporting
Know More