Generative AI & Agentic AI Engineering
From prompt to production agent — build systems that reason and act.
The flagship programme. You will go past chatbot demos and build what companies are actually hiring for: retrieval systems that cite their sources, multi-step agents that use tools safely, and evaluation pipelines that keep them honest in production.
- Duration
- 20 weeks
- Weekly effort
- 10–12 hrs/week
- Level
- Intermediate
- Format
- Live cohort + build labs
- Mode
- Classroom in Marathahalli & Live online
- Typical range
- ₹12–38 LPA
Outcomes
What you will be able to do
Not topics covered — capabilities you can demonstrate in an interview and on day one of the job.
- Explain transformer and attention mechanics well enough to debug model behaviour
- Ship a production RAG system with hybrid retrieval, reranking and grounded citations
- Design multi-agent systems with planning, memory, tool use and human-in-the-loop control
- Fine-tune open-weight models with LoRA/QLoRA and know when fine-tuning is the wrong answer
- Build the eval and observability layer that keeps an AI feature safe to release
Curriculum
6 modules over 20 weeks
Every module ends in working code that is reviewed by a mentor before you move on.
01Foundations of Modern LLMs
Weeks 1–3
Foundations of Modern LLMs
Weeks 1–3
- Tokenisation, embeddings and what the model is really doing
- Transformer architecture and attention, built from scratch in PyTorch
- Pretraining, instruction tuning, RLHF and why models behave the way they do
- Context windows, sampling parameters, and reading a model card critically
02Prompt & Context Engineering
Weeks 4–5
Prompt & Context Engineering
Weeks 4–5
- Structured prompting, few-shot design and decomposition patterns
- Reasoning techniques: chain-of-thought, self-consistency, reflection
- Context window budgeting, compaction and long-document strategies
- Prompt injection, jailbreaks and defensive prompt design
03Retrieval-Augmented Generation
Weeks 6–9
Retrieval-Augmented Generation
Weeks 6–9
- Document parsing, chunking strategy and metadata design
- Embedding model selection and vector database internals (HNSW, IVF)
- Hybrid search with BM25, reranking, and query rewriting
- Grounded answers, citation enforcement and hallucination reduction
- GraphRAG and structured-data retrieval over SQL
04Agentic AI Systems
Weeks 10–14
Agentic AI Systems
Weeks 10–14
- Agent architectures: ReAct, planner-executor, supervisor and swarm patterns
- Tool design and Model Context Protocol servers
- Short-term, episodic and long-term memory design
- Multi-agent orchestration with LangGraph: state, checkpoints, recovery
- Human-in-the-loop approval, sandboxing and blast-radius control
- Computer-use and browser agents, and their failure modes
05Fine-Tuning & Open Models
Weeks 15–17
Fine-Tuning & Open Models
Weeks 15–17
- When to prompt, when to retrieve, when to fine-tune — a decision framework
- Dataset construction, cleaning and synthetic data generation
- LoRA, QLoRA and parameter-efficient fine-tuning on a single GPU
- Preference tuning with DPO
- Quantisation, distillation and serving with vLLM
06Evaluation, Safety & Production
Weeks 18–20
Evaluation, Safety & Production
Weeks 18–20
- Building eval sets; Ragas, LLM-as-judge and human review pipelines
- Tracing, cost tracking and regression testing for non-deterministic systems
- Guardrails, content safety, PII redaction and audit trails
- Deployment, streaming, scaling and capstone demo day
Capstone
What you walk out with
Portfolio work substantial enough to carry a technical interview on its own.
A production-grade agentic assistant over a real document corpus, with tool use, citations, evals and a public demo
A fine-tuned open-weight model that beats a general-purpose baseline on a domain task you choose
Tools & stack
Roles this leads to
- Generative AI Engineer
- AI / Agent Engineer
- LLM Application Developer
- AI Research Engineer
- Conversational AI Lead
Who it is for
- Developers who want to move into AI engineering
- Data scientists shifting from classical ML to LLM systems
- Technical founders building an AI product
Prerequisites
- Solid Python fundamentals
- Basic understanding of APIs and Git
- No deep learning background required — we build it up
Enrol
Get the full syllabus for Generative & Agentic AI
Request the detailed module-by-module syllabus, current batch dates and fee structure. A counsellor will also check whether this is the right programme for your background.
Also consider