We started by teaching.
That never stopped.
AI-Shala began in 2021 as one-to-one mentorship in AI and robotics. The products came later, and they came out of the teaching — not the other way around.
Every faculty member we taught alongside described the same week. Three hundred scripts. A deadline. And somewhere in that pile, thirty students who would have understood the subject if someone had told them precisely where their reasoning broke — and no hours left in which to do it.
Meanwhile, on the research side, we were working on a question that sounded unrelated: how do you build a system whose decisions a person can interrogate and overrule? In robotics that question is about safety. It turns out to be the same question in assessment, where a mark that cannot be explained is a mark that cannot be defended.
GunanQ is what our answer looked like once a real examination cell had to rely on it. Not an auto-grader — an instrument that does the mechanical work and then defers, completely, to the person whose name goes on the result.
We kept the teaching because it is where the problems come from. The students in our programmes are the reason we know what is actually broken, and several of them are now co-authors on the work that fixes it.
A deeptech lab that ships products, publishes research, and teaches the students who will do both better than we did.
- Founded
- 2021
- Based in
- India
- Products
- 2
- Programmes
- 3
Four things, and we mean them.
A human holds the pen
We do not build systems that take a decision away from the person accountable for it. This is a constraint on the architecture, not a preference we could be talked out of.
Say the true thing
We publish who our programmes are not for. We do not promise ranks or acceptances. It costs us enrolments and it is worth it.
Research earns the product
Every product here started as a question we could not find a satisfying answer to. If we cannot explain why it works, we are not ready to sell it.
Attention is the scarce thing
Content has been free for a decade. Everything we build or teach is ultimately about making real attention reach more students.
Five years, in order.
- 2021
AI-Shala begins as mentorship
A handful of students, one-to-one, in AI and robotics. No product, no plan — just the observation that the students who moved fastest were the ones with someone to ask.
- 2023
Research cohorts and first papers
The mentorship formalises into a research programme. Students start submitting to international venues, with several papers accepted.
- 2024
Physical AI on real hardware
Two cohorts take learned policies from simulation onto real robots. The tooling built along the way becomes the basis of SimRoboX.
- 2025
GunanQ ships
The interpretability work meets the problem every faculty member we taught kept describing: 300 scripts, 48 hours, and no time left to teach.
- 2026
A company, not a product
GunanQ moves to its own home at gunanq.com. SimRoboX enters private pilots. The programmes keep running, because they are where the questions come from.
Small team.
Named mentors.
Every student in a programme knows exactly who they are working with before they enrol. We are putting full profiles up here — in the meantime, you meet your mentor on the first call, not after payment.
Engineering
Builds and runs the products faculty depend on during exam season.
Research
Physical AI, interpretability and the questions underneath both products.
Education
Mentors who are working engineers and researchers, teaching because they are good at it.
Operations
The people who make sure a cohort starts on time and a demo actually happens.
Two doors in.
Institutions: bring a real question paper and we will grade it in front of you. Students: tell us where you are trying to get to.