A live, instructor-led course on how modern AI actually works: retrieval, evaluation, observability and multi-agent automation. Taught from first principles, for people who are not programmers.
You've probably tried one. Week 1: install Python. Week 2: learn FastAPI. Week 3: copy a LangChain tutorial. Week 4: you have a chatbot that works on the instructor's laptop. Then someone asks "why did it give the wrong answer?" — and you have nothing.
AI can write code. AI cannot tell you what to build, why, or whether it's working. That is the job.
Four kinds of people get very different things out of the same 40 sessions.
PMs, analysts, consultants, ops leads, founders. You'll be able to design, scope, review and confidently lead AI projects.
One clear, ordered mental model — instead of two hundred scattered YouTube videos that never connect.
Three portfolio projects and the vocabulary to survive an AI system-design interview.
The AI-specific layer you're missing — retrieval, evaluation, observability, agents. Harder homework variants included.
Python syntax · FastAPI · web development · DevOps · software engineering fundamentals. If you want to learn programming, this is the wrong course.
Eight parts, nineteen modules, three projects. Click any module to see what's inside.
Most courses skip this entirely. It is why most students can never debug their own systems.
The goal here is to reduce awe. By the end, LLMs should feel unremarkable and predictable.
From naive to advanced. Every component, every trade-off, every failure mode.
A full part of the course, not a closing slide. This is what separates a demo from a product.
Taught deliberately late — so you see frameworks as convenience over concepts you already understand, not as magic.
We open with the honest position: most "multi-agent" systems should have been one agent. Then we teach the five cases where it's genuinely right.
The other two are automations — because that is what companies actually pay for.
A production-shaped RAG assistant over a private document library. Full ingestion pipeline, hybrid retrieval with reranking, citations, a real RAGAS evaluation suite, live tracing in Langfuse, and a cost dashboard.
Raw input goes in — a mailbox of invoices, or a folder of dropped PDFs. The system parses them, extracts structured fields, validates against business rules, enriches from a lookup, decides the route, then acts: writes to a database, sends a notification, or flags an exception for a human.
This is the strongest piece in your portfolio. It's the one that gets you hired.
A topic goes in on a schedule. A planner agent decomposes it, parallel research agents gather from the web and your internal documents, an analyst agent synthesises, a writer agent drafts, a critic agent reviews — a formatted report comes out.
Live, instructor-led. Every Saturday and Sunday, 2 hours each day — designed so you can do this alongside a full-time job or final-year college. Sessions are live-only; the complete written material for every module is yours to keep.
Recap + hook — a real broken system you can't yet explain
Intuition — the analogy, with no jargon at all
Concept + architecture — drawn live, box by box, in the order data actually flows
Break
Trade-offs and failure modes — you diagnose a broken system
Live AI-assisted coding demo — including watching it get things wrong
Check questions + design homework
Your concept knowledge is what catches the bug. That is the entire point of the course.
Nobody else grounds RAG in classical NLP. It's precisely why our students can debug retrieval when others can only guess.
The concept depth is engineer-level. We remove the syntax barrier, not the substance.
Most courses never teach you how to prove your system actually works. Here it's a quarter of the syllabus.
Two of three projects have no chat interface at all — matching what businesses actually buy.
Which model for which role, at what temperature, at what cost. A topic almost entirely absent from the market.
You learn what breaks and the symptom it shows — not just what works in a demo.
We'll tell you when a single agent beats a multi-agent system, and when a plain workflow beats both.
Complete written notes, diagrams, cheat sheets and interactive explainers for every module.
Yes — the course was designed for you specifically. Every technical term is defined the first time it appears. Code is shown only for you to read and recognise, never to write from scratch.
Yes, working with an AI coding assistant — which is how a large share of professional AI development is now done. You'll be the architect and the reviewer. The assistant types.
No. We don't train models, do statistics, or build neural networks from scratch. This is AI Engineering — building applications on top of existing models. It's a different, and currently more in-demand, skill.
The Python and web-development parts aren't there because you already know them. Parts 3, 4 and 6 — RAG, evaluation and observability, multi-agent systems — go to a depth most engineers haven't reached. Engineers get harder "go deeper" homework variants.
No. The entire course runs on free tiers and models running locally on your own machine.
No — sessions are live-only, by design. It keeps the whole cohort moving together, keeps the room genuinely engaged, and means your questions get answered in the moment rather than shouted at a video. If you do miss one: the complete written material for every module is yours, each session opens with a recap of the last, and the group channel is there for catching up.
You'll get something better: three documented, working projects with architecture docs and evaluation reports. That is what a hiring manager actually asks to see.
Because the short version is the one that leaves you unable to debug your own system. Depth is the product.
This is the founding cohort. The price goes up for every cohort after this one.
Working professionals · career switchers
Valid college ID required
That's ₹75 per hour of live, small-batch instruction — before counting the written material and project reviews.
The field is full of people who can run a tutorial, and nobody who can explain why the answer
was wrong.
This course makes you the second kind of person.
— Abhishek Mane · AI Engineer
Have questions about the syllabus, batch timings, or 1:1 career mentorship? Connect directly with the instructor below.
