Generative AI Course in Pune

Day 1: Foundations of LLMs & Prompt Engineering

• Session 1: Fundamentals of LLMs
  • Understand how LLMs work, including tokenization and embeddings
  • Explore how models like GPT interpret and generate text
  • Project: Visualize and compare embeddings
• Session 2: Prompt Engineering in Practice
  • Learn how to structure and format prompts effectively
  • Understand chat roles: system, user, assistant
  • Design multi-turn prompts and chaining logic
  • Project: Create a multi-turn chatbot
• Session 3: Prompt Engineering with OpenAI API
  • Apply prompting techniques for summarization, classification, and text transformation
  • Use OpenAI API to test and iterate prompts programmatically
  • Project: Implement and compare zero-, one-, and few-shot prompts for:
    • Text summarization
    • Sentiment detection
    • Email drafting
• Session 4: Building Systems using Generative AI
  • Apply prompt engineering to real-world text tasks
  • Break complex tasks into multi-stage prompt pipelines
  • Evaluate LLM outputs for safety, accuracy, and relevance
  • Project: Build a customer-support agent that flags, retrieves, and answers user queries

Day 2: LangChain & Agent Systems

• Session 5: Application Development using LangChain
  • Understand the LangChain framework
  • Use Retrieval-Augmented Generation (RAG) with your own data
  • Build chatbots using structured chaining logic
  • Project: Create a document Q&A assistant using LangChain and PDFs
• Session 6: Multi-Agent Systems with AutoGen
  • Understand Agentic AI using Microsoft’s AutoGen framework
  • Orchestrate multi-agent workflows with memory and tools
  • Compare agent patterns:
    • Coordinator–Worker
    • Role-Based Agents
    • Tool-Invoking Agents
  • Explore real-world automation use cases
  • Optional: Compare AutoGen with CrewAI
  • Project: Build a 3-agent AutoGen customer support system (intake, triage, escalation)
• Session 7: Agentic Design with LangGraph
  • Design structured agent workflows using LangGraph
  • Handle retries, errors, and human checkpoints
  • Combine LangGraph with AutoGen for robust pipelines
  • Project: Build a LangGraph + AutoGen hybrid agent with dynamic routing and retries
• Session 8: Reasoning with LLMs
  • Understand reasoning models and their use cases
  • Apply chain-of-thought and self-reflection prompting
  • Use reasoning for code debugging and vision tasks
  • Project: Debug code and solve image-based questions step-by-step using reasoning prompts

Day 3: Reasoning Models & Advanced Agents

• Session 9 & 10: Advanced Use Cases with AutoGen
  • Design collaborative multi-agent systems with planner–executor architecture
  • Integrate multiple LLMs, tools, and shared memory
  • Enable agents to critique, reason, and self-correct
  • Apply agent systems to HR, legal, and customer operations
  • Project: Build an AutoGen-based resume screening system with ranking and critique logic
• Session 11: Build with Anthropic – Computer Use
  • Understand Anthropic models and best practices
  • Implement prompt caching for latency and cost optimization
  • Build multimodal and tool-augmented workflows
  • Project: Create an AI assistant that opens files, summarizes content, and suggests edits
• Session 12: Reasoning + Agent Project Sprint
  • Apply agent collaboration in a real-world use case
  • Combine LangGraph workflows with AutoGen agents
  • Project: Build a “Virtual Analyst” that analyzes documents, answers queries, and summarizes trends

Day 4: Memory, Context Protocols & AI Safety

• Session 13: Long-Term Memory with LangGraph & AutoGen
  • Build agents with persistent and evolving memory
  • Enable personalization and session continuity
  • Project: Extend AutoGen agents with long-term adaptive memory
• Session 14: Model Context Protocol (MCP)
  • Understand MCP for standardized tool and data access
  • Connect multiple AI agents via shared context
  • Deploy MCP servers and clients
  • Project: Deploy an MCP server and connect mini-apps with shared user context
• Session 15: Guardrails for AI Systems
  • Identify common LLM failure modes
  • Validate and filter AI inputs and outputs
  • Integrate guardrails into RAG and agent-based systems
  • Project: Add guardrails to block unsafe, biased, or irrelevant responses
• Session 16: AI Project Best Practices – End-to-End Deployment
  • AI product lifecycle: scoping, prototyping, testing, deployment
  • Trade-offs: prompting vs finetuning, cost vs accuracy, latency vs UX
  • Finetuning fundamentals
  • Best practices for versioning, observability, feedback, and HITL
  • Evaluate AI systems for quality, safety, and robustness