Free
₹0
- First 5 lessons (Phase 1 W1–W5)
- 5 AI tutor messages / day
- Hindi + Hinglish content where authored
- No cohorts, no portfolio publish
For Class 9 → graduation · India
Learn ML, Generative AI, and Agentic AI from first principles — the maths that makes them work. You build every idea by hand (never just watch), with an AI tutor that gives hints, not answers.
No credit card. No ads. Under-16 accounts require parent consent.
Not sure where you stand? Take the 60-second AI Readiness Check →
78 weeks
6 phases, one path
234 lessons
every one hands-on
3 tracks
Class 9-10 / 11-12 / Graduation
Interactive
a live demo in every lesson
What you'll be able to do
Attention, transformers, diffusion: the real equations, worked by hand, not analogies.
From a single neuron to backpropagation to a working neural network you wrote yourself.
A reviewed capstone: a tool-using agent that plans and acts — deployed, not just demoed.
Linear algebra, calculus, probability — the foundation every model is built on.
Not a video archive. Not a homework grinder. A guided sequence where every minute earns a build.
Every concept asks you to guess first. Wrong guesses get a per-option explanation — your mistake is the teaching moment, not a dead end.
Each lesson ends with a small thing you build — a Python script, a model, a tool, an agent. By Week 9 you've written your own.
Nova / Coach / Mentor (changes by your age band) gives you nudges, never the full answer. Get unstuck without skipping the work.
Under-16 accounts gated by parent OTP. No DMs between students. Weekly progress emails to parents. DPDP-compliant.
Lessons surface the CBSE / ICSE chapters they cover. Helps with school math too — not just AI.
₹399/month or ₹2,000/year (saves 58%). First 5 lessons free, no card required.
Don't take our word for it
This is a real one — gradient descent, the algorithm that trains every model in the course. Press Step to roll the ball downhill; push the learning rate too high and watch it overshoot and blow up. No slides, no video — you move the sliders.
Set a learning rate, then Step downhill
Loss surface
Step 0
loss = 6.25
x = -2.5, gradient = -5 · One minimum at x = 0. Convex — descent always finds it.
Start with algebra; finish having shipped your own agentic AI capstone. Each phase builds on the last; nothing skipped.
Algebra, calculus, vectors, matrices, Python, Git.
Linear models, trees, ensembles, evaluation, real datasets.
Neural networks, backprop, CNNs, RNNs, transformers.
LLMs, diffusion, RAG, embeddings, multimodal.
Tools, planning, memory, MCP, multi-agent systems.
MLOps, serving, monitoring, your final project.
Tap any phase to see representative topics.
Three depth-tuned tracks (Class 9-10, Class 11-12, Graduation) — same curriculum, depth adjusted.
Five stages per lesson. Each stage is a few minutes. You can't skip the checks — that's on purpose.
Email + (if you're under 16) a parent OTP. Takes 2 minutes. No card needed.
Every concept: guess first, see the rule, lock it in with a quick check. Get hints from Nova if you're stuck.
A worked example walks through an expert's thinking step by step — not just the answer.
Write a 2-3 sentence explanation in your own words. The AI grades it against a rubric. That's how you know you got it.
5-6 question synthesis. Pass at 80% to mark the concept mastered. Then on to next week.
Why it works
Predict-then-reveal and build-don't-watch aren't our opinions. They're decades of learning-science research, built straight into every lesson.
Students in active-learning classes are about 1.5× less likely to fail than those sitting through lectures.
Freeman et al., 2014 — meta-analysis of 225 studies, PNAS
Why every lesson ends in something you build, not a video you watch.
Pulling an answer back out of memory cements it far better than re-reading ever does — the “testing effect.”
Roediger & Karpicke, 2006 — Psychological Science
Why we keep checking: quick checks, weekly boss checks, spaced revision.
You remember what you work out yourself better than what you’re simply shown — the “generation effect.”
Slamecka & Graf, 1978 — J. of Experimental Psychology
Why you predict and explain it back before we ever reveal the rule.
Motivation is a system, not a mood. These aren't decoration — each one is tuned to keep you learning week after week.
Show up daily; keep your streak alive. Miss a day and it resets — gentle pressure that builds the habit.
Every predict, check and build earns XP. Level up as concepts stick.
Milestones, board-exam wins and phase completions get their own badge.
End each week with a 5-question boss check. Pass at 80% to mark it mastered.
Join an anonymous peer cohort — compare progress and climb the leaderboard, no names shown.
Spaced-repetition nudges resurface older concepts right before you'd forget them.
Try the first 5 lessons for free. Subscribe when you're sure.
₹0
₹399per month
₹2,000per year
| Feature | Free | Premium |
|---|---|---|
| Lessons unlocked | First 5 (Phase 1, W1–W5) | All 234 across 78 weeks |
| AI-tutor messages / day | 5 | 100 |
| Interactive visual playground per lesson | In your 5 free lessons | In every lesson |
| Anonymous peer cohorts | — | ✓ |
| Portfolio publishing | — | ✓ |
| Completion certificate | — | ✓ |
All prices in INR, GST inclusive. See Refund Policy.
For parents
No account is created until you (the parent) approve via an OTP we send to your phone or email. You can revoke access anytime.
Every Monday: what your child learned last week, where they got stuck, how much screen time they spent. No surprises.
Under-16 accounts cannot message other learners. There are zero advertisements. We don't sell data — ever.
We follow the Digital Personal Data Protection Act, 2023. All policies in plain English on our /legal page.
Lessons surface the school-board chapter they cover. Studying for AI also studies for boards.
Cancel within 7 days of subscribing and get a full refund. After that, pro-rata for any outage on our side.
Why we built this
Most “learn AI” courses are videos you nod along to and forget. Math to Machine is the opposite: every concept makes you predict, build, and explain it back before you move on. That method — the same learning science behind the mastery loop — is what the whole platform is built around.
You'd be joining the founding cohort: the first students shaping the platform. Launch pricing is locked in, and your feedback directly steers what we build next.
Don't see your question? Email us.
Yes. Phase 1 starts with school-level algebra and Python. If your child can do Class 9 maths, they can start. The curriculum spans up to first-year engineering depth, so they grow with it.
Phase 1's foundations (algebra, calculus, statistics) directly overlap with Class 9-12 maths. We surface the relevant board chapters on every lesson. Phase 2+ is beyond school syllabi.
Almost any laptop or phone from the last five years works. The interactive concept demos run right in the browser. Python is installed once — Week 1 walks your child through it step by step, and the AI tutor helps decode any error. The heavier later phases use free cloud notebooks we point you to, so no expensive GPU is ever required.
Built-in AI tutor (Nova for Class 9-10, Coach for Class 11-12, Mentor for graduation years) gives hints — never the full answer. After 4 hints + reflection, it suggests asking a teacher or parent.
Yes. Account creation requires parent OTP. Under-16 accounts can't DM others, can't publish publicly without parent opt-in, and don't see ads (there are none). DPDP-Act compliant.
Yes. 7-day no-questions refund for new subscribers. After that, the subscription runs to its end and just doesn't renew. We don't auto-charge during this launch.
Yes. English is available throughout, and selected early weeks also have Hindi and Hinglish (Roman-script Hindi mixed with English technical terms) versions. Anything not yet translated shows in English automatically — per step — so your child is never blocked, and you'll see exactly how much of a lesson is translated. Hindi coverage expands over time.
Pause anytime. Their progress is saved. Come back when they're ready. The curriculum is self-paced — no batches, no deadlines.
See if the predict-then-reveal model works for your child. If it doesn't, you've lost nothing — no card needed.