Programme Content
CURRICULUM
Fifteen Modules, One Working System
PHASE 1Python, ML, deep learning, and embeddings 48 hrsPHASE 2Generative models and the LLM stack 33 hrsPHASE 3Retrieval and adaptation 21 hrsPHASE 4Agentic AI and multi-agent systems 15 hrsPHASE 5LLMOps, deployment, and the capstone 28+ hrsPython, ML, deep learning, and embeddings 48 hrs
M0 to M1 Foundations 27 hrs
Python recall, AI, ML, and deep learning, neural networks, CNNs, and backpropagation.You walk out able to: work fluently across the classical stack, so nothing in the LLM modules rests on a gap.
M2 to M3 NLP and Embeddings 21 hrs
Transformers, BERT, vector databases, semantic search, FAISS.You walk out able to: build with transformer architectures and run semantic retrieval over a vector store you set up yourself.
Generative models and the LLM stack 33 hrs
M4 to M5 Generative AI 22 hrs
GANs, VAEs, diffusion models, LLM architecture, open-source LLMs.You walk out able to: build across text, image, and multimodal generation, and work inside an open-source LLM rather than around it.
M6 to M7 Prompt Engineering and LangChain 11 hrs
Prompt strategies, structured JSON output, LangChain chains, memory, introduction to RAG.You walk out able to: enforce reliable structured output and chain memory-aware pipelines.
Retrieval and adaptation 21 hrs
M8 to M9 RAG Systems 15 hrs
Basic RAG, Advanced CRAG, Graph RAG on Neo4j, RAGAS evaluation.You walk out able to: ship production retrieval and prove it works with a real evaluation harness.
M10 Fine-Tuning LLMs 6 hrs
LoRA, QLoRA, instruction tuning, PEFT, LM Eval Harness.You walk out able to: fine-tune efficiently on constrained hardware and benchmark the result.
Agentic AI and multi-agent systems 15 hrs
M11 Agentic AI and Multi-Agent Systems 15 hrs
LangGraph, Microsoft AutoGen, CrewAI, Agentic RAG, multi-agent systems.You walk out able to: build agent teams that plan, delegate, and recover from their own failures.
LLMOps, deployment, and the capstone 28+ hrs
M12 to M14 LLMOps and Deployment 16 hrs
MLFlow, Docker, Kubernetes, FastAPI, LangSmith, low-code and vibe coding.You walk out able to: deploy, monitor, trace, and scale an AI system in production.
M15 Capstone 12+ hrs
End-to-end agentic build, deployment, and evaluation by IIT Roorkee faculty.You walk out able to: run one problem from architecture to a live, monitored system and defend every choice in it.
