The AI Fellowship.
Chapter four — and deliberately not for everybody.
This is where people go when the honest answer really is hardcore AI engineering. Three months of structured curriculum, live learning, industry mentors and real projects — for the few who want the deep end rather than a broader path. If you have not had a counselling session yet, take that first; it costs nothing and it is the better place to start.
Not certain this is your direction? Take free counselling first
Cohort 8 · September 8
Production AI, not toy demos
Every module ends with something running — RAG pipelines, agents, evals, deployment. Live projects, not exercises with a known answer.
Mentors from top labs
1:1 mentorship and code review with engineers who are shipping LLM systems in industry right now.
Direction, week by week
Your mentor helps you work out which specialisation actually suits you — and tells you honestly when something is not your strength.
Curriculum
Understand → Build → Apply → Prove
Four phases across three months. Each one ends with something running, and each builds directly on the last. Nothing here is theory for its own sake — if it will not show up in your work, it is not in the syllabus.
AI Foundations
Before you build with a model, you learn what a model actually is. Four weeks of first principles, taught by people who read the papers when they came out.
- Transformer architectures from scratch
- Prompt engineering & function calling
- Embeddings & vector stores
Semantic search over your own documents
ML & Model Building
How models actually learn — features, training, evaluation and the failure modes nobody mentions until your first model quietly goes wrong in production.
- Supervised learning end to end
- Feature engineering & data quality
- Evaluation, error analysis & drift
A trained, evaluated model on real data
Building with LLMs & Fine-tuning
Retrieval that holds up on messy data, agents that recover from failure, and the part almost no course teaches — making a model yours, then making it cheap and safe enough to run.
- RAG systems that actually work
- Agents, tools & multi-step reasoning
- LoRA / QLoRA fine-tuning, serving & guardrails
A fine-tuned, retrieval-backed assistant
Capstone & Career
You build something real with an industry partner, then turn three months of work into a portfolio you can walk anyone through — line by line.
- Capstone with an industry partner
- System design for AI products
- Portfolio and GitHub review
A capstone shipped with a real company
The rhythm
What six hours a week actually looks like.
Structured enough to keep you moving, loose enough to fit around a job. This is the rhythm, every week, for three months.
Live concept session
90 minutes with a practitioner, not a recording.
Build time
Your own hours. The project is the homework.
1:1 mentorship
A mentor reads your work and tells you the truth about it.
Cohort demo
You present. Peers question. Everyone learns.
Outcomes
By the end, you should have all of this.
Not a certificate. A set of things you can point at, demonstrate and talk about.
“I came from a backend role with zero ML. Three months later I was fine-tuning models and shipping RAG systems. The mentors made the difference.”
“The Fellowship is genuinely production-grade. My capstone became a product feature at my current company. Best decision of my career.”
What you leave with
Three months, and three things that stay with you.
Most programmes hand you a certificate and wish you luck. We would rather you left with things that keep working long after the cohort ends.
Three things you built
Not cloned, not followed along — designed, broken, instrumented and fixed by you, with the commit history to show it.
A specialisation that fits
Three months is long enough to find out what you are actually good at. Your mentor helps you name it.
People who know your work
A cohort and a mentor who have read your work line by line — and who keep answering long after the cohort ends.
Apply to the next cohort.
Cohorts are deliberately small — a 1:10 mentor ratio only works that way. The Fellowship is ₹50,000 + GST for the three months. Secure your seat below, or talk to us first if you are unsure it is the right next step.
AI Fellowship — 3 months
Two live learning sessions a week and roughly 150 hours of project-based work.
- Programme fee
- ₹50,000
- GST (18%)
- ₹9,000
- Total payable
- ₹59,000
Payments are handled by Razorpay. Card details never touch this site.
COHORT 8 · STARTS SEPTEMBER 8
Who we’re looking for
- Comfortable with Python (or willing to become so)
- 1+ year of any engineering / analytical work — or a strong portfolio
- Two live sessions a week for 3 months — non-negotiable
- A real intent to build a career in AI
Not there yet? Start with a free path and apply to the next cohort.
Browse the free pathsBefore you apply
The honest answers.
Anything we haven’t covered? Ask a human.
Ask about the FellowshipNo. Most fellows arrive from backend, frontend, data or analytics roles with no ML experience at all. What we do need is comfort with Python — or a genuine willingness to get comfortable fast — and at least a year of engineering or analytical work behind you.
Yes, and most fellows do. Two live sessions a week for 3 months, with the rest project-based — around 150 hours in total, fully remote, scheduled around working hours. We are honest about one thing though: those hours are not negotiable, and the people who protect them are the ones who finish with something real.
No, and we would be suspicious of anyone who did. We are an education and career guidance platform, not a placement agency. What we commit to is that you will understand this work deeply, have three projects you built yourself, and leave with a clear, honest read on where you stand and what to do next.
Deliberately small. A 1:10 mentor ratio only works if we keep the cohort tight, so seats are genuinely limited — this is real capacity, not a countdown gimmick. We would rather turn someone away than dilute the mentorship everybody came for.
Talk to your mentor early — that is what the 1:1s are for. Sessions are recorded, labs stay open, and in genuine cases we will defer you to the next cohort rather than let you finish half-taught.