Indian Institute of Technology Bombay

Agentic AI

Indian Institute of Technology Bombay

Other

About Programme

Introduction: The Certificate in Agentic AI is a five-month, online, hands-on certificate designed for professionals in Technology, Software Development, Data Science, Artificial Intelligence, and Machine Learning roles. Delivered by the reputed faculty at the Department of Computer Science and Engineering (CSE) at the Indian Institute of Technology Bombay, the certificate is intended for learners who want to move beyond prompting and begin building autonomous AI agents that can reason, act, and collaborate to solve complex real-world problems. The curriculum is structured across four modules. It begins with essential AI foundations before advancing into core Agentic AI concepts, where learners are exposed to agent workflows, architectures, and memory frameworks such as RAG and MCP, while gaining hands-on experience with orchestration libraries including CrewAI and LangGraph. From there, learners will dive into advanced systems covering planning, reasoning, decision-making, multi-agent communication, and reinforcement learning. The final module prepares learners to deploy agentic systems responsibly at scale, with a focus on human-in-the-loop design, alignment, safety, monitoring, and security. This course is delivered through a mix of live, online, interactive sessions from IIT Bombay faculty, guided labs, and hands-on projects focussed on building agentic systems. Upon successful completion, learners will earn a Certificate of Completion from IIT Bombay and develop the skills required to build intelligent AI agents that go beyond chat-based applications to deliver real business value.

Programme Content

Module 1: Foundations of Agentic Systems This module equips learners with the technical foundation needed to build and deploy intelligent agents. Starting with a Python refresher focused on concurrency and API interactions, the module moves into the key building blocks of AI and Large Language Models, including tokenisation, context windows, and parameter tuning. It concludes with practical approaches to prompt engineering, covering persona design, structured outputs, and control over model behaviour. Topics Covered: Python Refresher for Agentic AI AsyncIO and API Interactions Introduction to AI and LLMs Transformer Architecture and Tokenisation Temperature, Top-P, and Sampling Prompt Engineering: Zero/Few-shot, System Prompts, Structured Output Module 2: Agentic AI Fundamentals This module introduces the core workflows and infrastructure behind agentic systems. Learners will understand how agents perceive, plan, and act in cycles, use tools and memory eectively, and route tasks based on user intent. This module also introduces the Model Context Protocol (MCP) for exposing local data to agents and explores two leading orchestration frameworks (LangGraph and CrewAI) for designing flexible, modular agent pipelines. Topics Covered: Agent Workflows and Architecture Tool Use and Function Calling Router Patterns for Task Delegation Memory and Knowledge Retrieval (RAG) Vector Databases and GraphRAG Model Context Protocol (MCP) Orchestration with LangGraph and CrewAI Module 3: Advanced Agentic Systems This module focuses on building more intelligent and adaptive agents. Learners will implement step-by-step reasoning (CoT), interleaved tool use (ReAct), and reflection mechanisms that help agents learn from past actions. Reinforcement learning techniques and prompt optimisation using DSPy are introduced, along with methods for coordinating multiple agents through voting, debate, and role handos. Topics Covered: Planning, Reasoning, and Decision-Making (CoT, ReAct, Plan-and-Solve) Self-Reflection and Trajectory Optimisation Dataset Preparation for Fine-Tuning Reinforcement Learning and Reward Training Best-of-N Sampling and Feedback Loops DSPy and Auto-Prompting Multi-Agent Communication and Coordination Module 4: AI Agents in the Real World The final module prepares learners to build deployable, safe, and trustworthy agentic systems. Key practices in monitoring, human-in-the-loop workflows, and prompt security are covered, alongside strategies for scaling and exposing agents via modern deployment stacks. Learners will explore trade-os in autonomy, ethical design, and best practices for testing and maintaining agent behaviour in production.

Programme Benefits

HANDS ON FACILITY:
The "Analyst" Agent Build a single agent that can search the web and summarise financial news. Concepts Covered: Tool Usage (Search API), Prompt Engineering, and Structured Output. Customer Support Agent A RAG-based customer support chatbot to address the frequently asked questions (FAQs) and escalate to a human operator if a customer needs specialised assistance. Concepts Covered: Natural Language Processing (NLP), LLM Integration, Agentic Frameworks (LangChain), Retrieval Augmented Generation (RAG), Human-in-the-loop (For Escalated Cases). “Software Engineering” Team A multi-agent system that takes a feature requirement document, handles the end-to-end software development life cycle, and returns a fully functional software code along with documentation Concepts Covered: Software Design, Code execution (sandboxing), Error Handling (Self-Correction if Code Fails), ReAct Pattern, Software Testing and Integration, Multi-Agent Collaboration.

Testimonials

"undefined"

Anonymous

Other programs in this subject area you might find useful

Same topic, Similar duration - Broader exploration across all institutes

Contact us for the further details

Speak with an Advisor

  1. Pedagogy (case method, experiential learning, coaching, simulations)
  2. Target Audience
  3. Programme Objective
  4. Brochure

Prashansa Uttam

Programme Advisor

+91 9403890085[email protected]Mon – Fri, 9am – 5pm IST

Indian Institute of Technology Bombay

Executive Education Office

NA
NA
NA
https://www.iitb.ac.in
Mumbai, India

Tell us about your program enquiry

Fill out the form below and our team will get back to you within 24 hours.

Agentic AI | IIT Bombay