AI engineering · Open · 5–6 months · 22 weeks
AI training — live online from anywhere.
AI Engineer
22-week AI Engineer path for 0–6 years: Python, SQL, ML evals, LLM APIs, RAG, then agents with tools and guardrails. Same length as Java/Python.
Early bird till 30 Sep
Live online
₹31,999/-₹15,999/-
Offer valid until 30 September 2026
Audience
Who it is for
0–6 years · Python helps
Freshers through about 6 years who want to ship AI features — RAG, tools, and agents. If you have never written Python, counselling may put you on Python Full Stack first or extra hours in weeks 1–5. If you already ship APIs (3–6 years), we compress Python/ML and spend more time on agents, evals, and production.
A tool-using agent you can demo (plan → tools → memory → traces), a RAG service with citations, and a FastAPI + Docker deploy. Roles this maps to: AI Engineer, GenAI / LLM Engineer, Python engineer on AI apps.
Roles this maps to
Week by week
Roadmap & syllabus
Live online
- 1
Weeks 1–3
Python for AI engineers
- Types, functions, packages, virtualenv, debugging
- GitHub from week 1: commits, PRs, README habit
- HTTP APIs, JSON, pydantic models
- 3–6 years: skip-ahead labs in the same weeks
- 2
Weeks 4–5
Data the agent will use
- SQL joins, filters, and the queries RAG still needs
- Pandas enough to clean a corpus and spot bad rows
- Files, chunking ideas, and “what is the source of truth”
- 3
Weeks 6–8
ML evals you can explain
- Train/test, leakage, baselines in scikit-learn
- Precision, recall, and why a demo can lie
- One tabular project so you can talk metrics in interviews
- 4
Weeks 9–11
LLM apps that return structure
- Chat, system prompts, temperature, token cost
- Structured output and function/tool schemas
- Embeddings and similarity search
- Safety: PII, prompt injection as a real failure mode
- 5
Weeks 12–14
RAG that cites sources
- Chunking, retrieval, rerank as an idea
- Citations, empty-hit behaviour, eval set of questions
- A docs Q&A service with a README a hiring manager can run
- 6
Weeks 15–18
Agents — plan, tools, memory
- Tool calling: search, SQL, HTTP, file write — with allow-lists
- Multi-step graphs (LangGraph or equivalent): loops, retries, human gate
- Short-term vs long-term memory; when an agent should stop
- Traces, cost caps, and a failure report you can show in an interview
- Build one domain agent (support, ops, or research) — this is the job skill
- 7
Weeks 19–20
Ship the AI service
- FastAPI around the agent/RAG
- Docker image, env secrets, basic rate limits
- Logging and traces; what to say in a production incident
- 8
Weeks 21–22
Capstone + interviews
- End-to-end agent demo + architecture sketch + README
- Python + SQL + “how would you eval this agent” mocks
- Resume mapped to AI Engineer / GenAI JDs
Build
Projects
- Evals notebook. Tabular baseline so you can explain metrics and leakage.
- RAG over your docs. Chunk, retrieve, answer with citations, and a small question-eval set.
- Tool-using agent. Plan → tools (search/SQL/API) → memory → stop. Traces and a cost cap. This is the capstone.
- FastAPI + Docker. Expose the agent or RAG as a service with env secrets and a README a hiring manager can run.
Intake
Batch
- Live online
- Next intake: 1 November 2026
- Adults; 0–6 years in software, data, or support — counselling sets the pace
- Python basics, or finish Python Full Stack / extra early weeks
- Laptop with 8 GB RAM; class plus about 2 hours of labs a day
You get
Included
- Batch seat
- Pace split on day 1: 0–2 vs 3–6 years
- Recordings after every class
- Labs, traces review, and mock interviews
- Resume + GitHub mapped to AI Engineer JDs