A hackathon in Chennai for college students. Build something that helps people learn.
The problem statement
A classroom has one teacher and dozens of students, each stuck in a different place. Families who can afford it pay for a private tutor. Your challenge: build an AI that gives every student that kind of attention, one to one, at their own pace.
- One to one
- One student, one tutor. Nobody waits for the class to catch up.
- Self-paced
- It slows down, speeds up and goes back when the student needs it, not when the syllabus says.
- Adaptive
- What it teaches next depends on what this student actually knows.
BrainBack builds AI tutoring for school students. The six students below are invented.
27%
of Indian school students were taking private coaching in 2025. In cities it was 30.7%. Families already pay to fill the gap.
NSS 80th round, Education 2025 (PIB)27%
of rural Std III children could read a Std II text in 2024: a book for the grade below. The gaps compound.
ASER 2024 national findings (Pratham)
It hears.
Nila is stuck on Pythagoras and talking herself through it on the terrace. Mid-sentence she mutters, "wait, why?" A good tutor stops and answers. Most software makes her type.
The hard partPicking her voice out of traffic noise, and telling thinking aloud from asking a question.
It speaks.
Arun thinks in Tamil but his exam is in English, so he flips between both mid-sentence. A good tutor flips with him, calmly, at his speed.
The hard partMixed-language speech, and a voice that sounds patient instead of robotic.
It sees.
Meera's working is on paper, not on a screen. A good tutor looks at the page and finds the exact line where a sign slipped.
The hard partReading messy handwriting, and judging each step instead of only the final answer.
It thinks.
Farah is stuck on quadratics because something earlier never clicked. A good tutor goes back and finds it, instead of explaining the same thing again, louder.
The hard partKeeping a live picture of what each student knows, and choosing what to teach next.
It draws.
Karthik is on the platform with his phone. A good tutor sketches the triangle while it talks and points at the side it means.
The hard partDrawing diagrams live, in step with speech.
It checks.
Divya says x = 3. AI models make things up, so the tutor must not just guess whether she is right. It needs a way to know.
The hard partChecking answers against something the AI cannot talk its way around.
Six students, one city. Every one of them gets a tutor that is only theirs. That is the challenge.
What you build
A working demo of one slice.
You do not need to build the whole tutor. Build one piece of it, and make it real.
- 01
Pick a slice.
One of the six abilities above, one of the open questions below, or an idea we have not thought of.
- 02
Pick a real student.
One topic, one kind of student. Say who is stuck on what, and take them from stuck to unstuck.
- 03
Show it live.
A working demo beats a slide deck. How it is judged is in the details below.
Design for the real world
- Patchy internet
- Cheap phones
- Students who think in more than one language
- A child's data and privacy
Now go further
What is still open?
The tutor above is the bar. These are the questions nobody has finished. Pick one, or bring your own.
- 01
Offline?
A tutor that keeps teaching with no signal: on a phone, in a classroom, on a bus.
- 02
Every language?
Every Indian language, with the accents and the code-mixing real students actually speak.
- 03
Writes its own lessons?
Turn a syllabus into lessons that teach, with checks that keep every one of them correct.
- 04
Gets better by itself?
A tutor that notices how its own lessons land and improves, without waiting for a release.
- 05
What else?
The best idea in the room may be one we have not listed.
The details
Anything we have not confirmed yet says “To be announced”.
- Who can join
- College students
- City
- Chennai
- Dates
- To be announced
- Format
- To be announced
- Venue
- To be announced
- Team size
- To be announced
- Starter kit
- To be announced
- Prizes
- To be announced
- Judges
- To be announced
Get your team in.
One registration per team, from the team lead. Tell us who you are and why your team should be in.