Building AI Agent using Generative AI
Durations – 10 Days
Program Structure
- GenAI Agent Fundamentals
- LLM + Tools & APIs
- Memory & RAG
- Planning & Autonomy
- Multi-Agent Systems
- Advanced Multi-Agent Workflows
- Enterprise Automation Agents
- Testing, Evaluation and Safety
- Capstone Project
Prerequisites
- Basic programming knowledge (Python preferred)
- Understanding of fundamental AI/ML concepts (not mandatory but helpful)
- Familiarity with APIs and JSON formats
- Ability to work with VS Code / Jupyter Notebook
Program Outcomes
- Build foundational understanding of Generative AI and Agentic AI concepts.
- Enable learners to design, develop, and deploy intelligent agents powered by LLMs.
- Teach practical integration of GenAI with tools, APIs, and real-world workflows.
- Develop skills for creating autonomous and multi-agent systems with planning, reasoning, and automated actions.
AI Tools / Platform:
- Generative AI / LLM Platforms – OpenAI GPT models, Llama models, Google Gemini
- Agent Development Frameworks – LangChain, LlamaIndex, CrewAI, AutoGen, or similar multi-agent orchestration tools
- Vector Databases / RAG Components – FAISS, ChromaDB
- Programming & Development Tools – Python, Jupyter, VS Code
- API Integration Tools – REST APIs & Webhooks
- Version Control – Git & GitHub
Assessment – MCQ, Module Test
| Generative AI & Agentic AI Fundamentals | ||
| Introduction to Generative AI and LLM capabilities, What is Agentic AI? |
Characteristics of agentic systems Agent architecture (Reasoning, Tools, Memory, Planning) |
Types of agents: Reactive, Reasoning, Planning, Multi-Agent, Modular programming concepts, functions, scope & lifetime |
| Agent Reasoning & Cognitive Frameworks | ||
| Chain-of-Thought (CoT), ReAct (Reason + Act), Tree-of-Thoughts, Iterative reasoning and self-correction | Task decomposition and step-wise execution | Building reliable reasoning patterns for agents |
| Tools, Function Calling & API Integration | ||
| Purpose of tools in agent systems, Tool invocation workflows, Function calling (OpenAI / Claude / Gemini) | Designing tool schemas (inputs/outputs), Validation, error handling & fail-safes | Connecting agents to external systems (search, email, IoT, data APIs) |
| Tool Libraries & Multi-Tool Workflows | ||
| LangChain toolkits, LlamaIndex tools | Custom tool creation, Multi-step tool | Real-world tool orchestration |
| and query engines | pipelines, | patterns |
| Memory Systems for Agents | ||
| Why agents need memory, Short-term vs long-term memory | Summarization-based memory updates, Embedding-based memory, Conversation persistence, | Memory support in LangChain & LlamaIndex |
| Retrieval Augmented Generation (RAG) | ||
| Introduction to RAG, Embeddings, chunking & indexing | Vector databases: FAISS, Pinecone, Chroma, RAG pipeline architecture, | Building document-aware agents (Ask-Your-PDF, Ask-Your-Docs) |
| Advanced RAG & Knowledge-Driven Agents | ||
| Multi-document retrieval, Query transformation & reranking | Hybrid search (keyword + vector), Domain-knowledge augmentation | Designing enterprise knowledge assistants |
| Agent Planning, Autonomy & Execution | ||
| Planning techniques for agents, ReAct-based planning loops, | Autonomy vs human-in-loop control, Self-reflection and iterative improvement, | Long-horizon task execution workflows |
| Multi-Agent Systems & Orchestration | ||
| Multi-agent architectures, Planner–Worker–Critic roles, Inter-agent communication, | Collaboration frameworks (AutoGen, CrewAI, HF Swarm) | Multi-agent orchestration for complex tasks |
| Test, optimize and secure agents for real deployment | ||
| Agent evaluation metrics | Guardrails & safety layers | Hallucination reduction, reliability checks |
| Project | ||
| Project Building | Project Demo | Project Evaluation |
