Seven hands-on sessions, each ~3 hours. No code, no installs — just a phone or laptop, a curious mind, and Indian context throughout. Every session ends with an artifact you ship to your public profile.
A 15-minute onboarding before your real sessions begin. Meet AI Buddy — your AI tutor — and discover the most useful trick of the whole course. AI Buddy can help you with anything from your regular school day, in Hindi, before we even teach you what AI is.
AI Buddy speaks Hindi. Try it with your toughest school subject — photosynthesis, the Mughal empire, why the Indus river changed course. Whatever you ask in school, ask it here in your own words.
Learning objectives
- ›Open AI Buddy and ask your first question in your own words
- ›Learn the simplest rule of asking — be specific
AI is not coming in the future. It is already in five apps on your family's phone right now. Today we hunt for it. We make a poster. We give it to AI Buddy and we ask why each one is "AI" and not just "software".
We use only Indian apps and Indian examples — UPI fraud detection, Swiggy / Zomato food recommendations, Google Maps in Bengaluru traffic, IRCTC's autocomplete, your phone's Hindi keyboard. These are the AI systems used by 800 million Indians, every day, mostly invisibly.
Learning objectives
- ›Spot 5 apps on a family phone that use AI
- ›Describe in one sentence what each AI is predicting
- ›Build a one-page digital poster of your findings
- ›Submit the poster to your public profile
Lessons
- ·AI is hiding in your phone45m
- ·Audit your family phone45m
- ·Make your "AI in my phone" poster30m
Portfolio piece — AI in my phone — my poster
A one-page poster of 5 apps on a family phone where you found AI hiding. Drawn by hand or made digitally — either is fine. The five sentences matter more than the design.
Today you become an AI teacher. You will train an AI to tell apart 3 things you have at home — 3 dals, 3 toy cars, or 3 different hand signs. You will also discover the most important rule in all of AI — more examples means a smarter AI.
We use what's already in your kitchen — toor dal, urad dal, chana dal. If you don't have these at home, hand signs work just as well — fist, peace sign, thumbs up. The rule we discover today is the same rule UIDAI used to train Aadhaar to recognise hundreds of millions of fingerprints.
Learning objectives
- ›Train an AI image classifier from scratch using Teachable Machine
- ›Discover that more examples make the AI smarter
- ›Save your trained model and share it with AI Buddy
- ›Write 3 lines about what you built
Lessons
- ·What does it mean to teach a machine?45m
- ·Train your first AI45m
- ·More examples = smarter AI30m
Portfolio piece — My first trained AI
Your first piece of AI you can show off. A 3-class image classifier you trained yourself on Teachable Machine. The 30-example version, not the wobbly 5-example version.
Last session you trained an AI on 3 things in your kitchen. This session we go bigger — we train an AI on Indian visuals. You'll pick something that lives outside your home and teach the AI to recognise it from 30+ photos. Then you'll make a 30-second demo video.
Pick from real-world Indian categories — types of mango (Alphonso vs Dasheri vs Banganapalli), levels of fruit ripeness, kinds of cricket equipment (bat / pads / ball / glove), styles of mom's earrings (jhumka / hoop / stud), or two-wheelers on your street (Activa / Bullet / Hero Splendor). The 9-12 Builder track will do exactly this — but with Python, on bigger datasets. You'll already feel comfortable.
Learning objectives
- ›Train an AI image classifier on a richer Indian visual category
- ›Take 30+ photos per class with deliberate variety in light, angle, and distance
- ›Record a 30-second phone demo of your AI working
- ›Submit the demo + classifier link to your public profile
Lessons
- ·Why "seeing" is hard for AI45m
- ·Train your "world classifier"45m
- ·Make a 30-second demo video30m
Portfolio piece — My world classifier
Your second trained AI — but this time on real Indian visuals from outside your kitchen. Mango ripeness, cricket gear, two-wheelers, sweets, anything visual and Indian. 30+ photos per class, deliberate variety in lighting and angle, plus a 30-second demo video.
Today you meet a Google AI trained on 50 million doodles from people all over the world. You doodle 8 things. The AI guesses each one — sometimes wrong in funny ways. You'll learn the difference between *recognising patterns* and *understanding the thing*.
Try drawing things kids in India draw — a kurta, a cricket bat, a bindi, a thali, a lassi glass. The AI may know "shirt" but not "kurta". You'll discover what's missing from its training data — and why that matters.
Learning objectives
- ›Use Google Quick Draw and notice what AI gets right vs wrong
- ›Understand that AI sees shape patterns, not the actual object
- ›Pick your best 3 doodles plus one that the AI couldn't guess
- ›Submit them to your public profile
Lessons
- ·Play Quick Draw — meet a 50-million-example AI45m
- ·How does it know? (without "knowing" anything)45m
- ·Your sketch wall30m
Portfolio piece — My sketch wall — 3 wins, 1 fail
A 2x2 grid of 4 doodles you made in Quick Draw — three the AI guessed right, one it couldn't. Each with a one-line caption explaining what you drew, what the AI said, and why you think it landed that way.
This session is a game. Your job is to break the AI you built in Session 3 — on purpose. Why does it fail under different lighting? What happens when you photograph a brand-new colour of mango? Whoever finds the most ways to break it makes a "model report card" listing what their AI is bad at.
This is the same skill the team at the Reserve Bank of India uses to test UPI fraud-detection AIs — they try to fool the AI before real fraudsters do. Your model report card today is a baby version of what real engineers call "adversarial testing."
Learning objectives
- ›Find at least 5 ways your Session 3 classifier fails
- ›Explain why each failure happens — usually because of training data
- ›Submit a "model report card" listing the failures
- ›Understand that finding flaws is a skill, not a problem
Lessons
- ·Why finding AI mistakes is a skill45m
- ·Try 10 ways to fool your AI45m
- ·Write your AI's report card30m
Portfolio piece — My AI's report card
A list of at least 5 ways you broke your own Session 3 classifier — what you tried, what the AI said, why it failed, what would fix it. The artifact that proves you can find flaws, not just build things.
This session you pick ONE small real problem from your own family life — and you train an AI to help. A lunchbox classifier, a ripe-mango detector, a recogniser for dad's handwriting. Build it. Test it on real data. Write down what works and what doesn't.
The point is to pick something *true to your family's actual life*. A kid in a Mumbai high-rise will pick something different from a kid in a Pune chawl who'll pick something different from a kid in a Bareilly farm. Whatever you pick, the recipe is the same one you've been using since Session 2.
Learning objectives
- ›Pick a real personal problem worth ~30 photos
- ›Train a 3-class classifier on that problem
- ›Test it on real data and document what works
- ›Submit the AI plus a 2-line "what works / what breaks" note
Lessons
- ·Pick your "AI for my family" problem45m
- ·Build it45m
- ·What works, what breaks30m
Portfolio piece — My family AI
An AI you trained to help with a real problem from your family's life. A lunchbox classifier, a ripe-mango detector, a handwriting recogniser, or anything similar that you actually find useful at home. 30+ photos per class, real-world testing in different conditions, plus an honest "what works / what breaks" note.
The final session of AIDE Explorer. You've built 4+ AIs across the term. Today you pick one — your proudest — and polish it for the world. Better description, better screenshot, share with three people. Walk away knowing you trained AIs at age 12.
This is the last item that lives on your public profile at aide.vcbeyond.org/p/<slug>. The classmates, family, and teachers you share that link with will see this AI front and centre. Make it the one you want them to see.
Learning objectives
- ›Pick your proudest AI from Sessions 2, 3, 4, or 6
- ›Polish its title, description, and screenshot/video
- ›Share with at least 3 real people
- ›Submit to your profile as the Showcase project
Lessons
- ·Pick your proudest AI45m
- ·Polish your Showcase45m
- ·Showcase day — share with three people30m
Portfolio piece — My AIDE Explorer Showcase
The capstone of AIDE Explorer. Your proudest AI from the term, polished — better title, sharper 3-line description, one strong screenshot or video. The big card on your public profile that's seen first when someone visits aide.vcbeyond.org/p/<your slug>.