Skip to content
Deeptech education · India

Architecting the
Intelligence of Tomorrow.

Products that give teachers their week back. Programmes that put students in front of real research. And a lab that ships both.

Products built for real institutions Mentorship from a named person Research published, not promised Human-in-the-loop by design Your data is not our training set Built in India, used on real campuses Products built for real institutions Mentorship from a named person Research published, not promised Human-in-the-loop by design Your data is not our training set Built in India, used on real campuses
What AI-Shala is

One lab, three ways of
showing up.

We build products for the classroom, teach the students who will build the next ones, and publish the research underneath both. Each feeds the others — that is the whole thesis.

Flagship product Live

Grading is where
teaching goes
to die.

Not because faculty stopped caring — because 300 scripts and 48 hours leave no room for the one thing that actually teaches: telling a student exactly where their thinking broke.

GunanQ gives that room back. It measures every script against your rubric, then stops and waits. Nothing reaches a student until a human presses Finalize.

GunanQ
Products

Everything we've built,
and what's next.

All products
RUBRIC 2/2 2/2 1/2 2/2 Proposed total 7 / 8 Awaiting faculty sign-off
Assessment AI Flagship

GunanQ

Rubric-grounded assessment that carries the mechanical weight of grading, then stops. AI proposes, faculty decides, and nothing reaches a student until a human presses Finalize.

  • Rubrics
  • Handwriting OCR
  • Human-in-the-loop

Institution pricing

Per cohort · annual licence

Details
TODO IN REVIEW DONE PR attached every move timestamped · open work carries forward WEEK 34 · COUNTED Carried in 7 Added 12 In review 3 Blocked 1 Done 11 Nothing typed in twice.
Team Operations Live

TeamPulse

A task board, wiki, OKRs, attendance and weekly reporting in one application — where the weekly report is counted from the board rather than typed into a document, so a status update cannot disagree with reality.

  • Task board
  • Weekly reporting
  • Wiki

Access on request

Per team

Details
4 participants Priya Arun Meera You Recording no bot joined · nothing to admit MINUTES · READY Summary Decisions Action items Priya Fri Arun Mon You Wed
Meetings Live

StorySoFar

One click in the call you are already in, and structured minutes are waiting when it ends — summary, decisions and action items with owners. No bot joins, nothing to admit, and your own voice is in the notes too.

  • Google Meet
  • Live transcription
  • Action items

Access on request

Chrome · Google Meet

Details
QUESTION 17 not just wrong — wrong in a known way MISCONCEPTION Applies the safety check without testing Need seen in 7 responses QUEUE FOR TOMORROW
Assessment AI Live

Atlas APM

A preparation engine for high-stakes exams. It names the specific misconception behind a wrong answer, schedules the day around it, and keeps what you have learned from decaying before exam day.

  • Diagnostics
  • Adaptive scheduling
  • Spaced repetition

Access on request

Per cohort · institutions

Details
the shelf is empty now EPISODIC MEMORY 09:42 · kitchen 11:15 · store room yesterday · bench last week · lab you moved it at 11:15
Physical AI Coming soon

Memory for Embodied AI

Persistent, grounded memory for embodied agents. In development — more when there is something worth showing.

  • Episodic recall
  • Spatial grounding
  • Sim and real

Coming soon

Coming soon
EPISODE 4.2k sim real
Simulation Coming soon

SimRoboX

A high-fidelity physics simulator for reinforcement learning agents. Build complex 3D environments, train policies at scale, and transfer them to real hardware without breaking anything expensive.

  • Gym-compatible
  • Domain randomization
  • Multi-agent

Early access

Free for academic labs

Details
Programs

Mentorship, the way
it should have been.

Small cohorts, named mentors, and someone who reads your work every week.

Compare programmes

GATE Preparation

AIR 11 · GATE CS

Solve problems in front of someone who can see how you think.

Format
Live, small cohort
Duration
6–11 months
From
₹24,000
Next cohort
Rolling admissions
Explore programme

Research Mentorship

3+ papers at top-tier AI venues

A real problem, a supervisor who reads your drafts, a submitted paper.

Format
1:1 + lab group
Duration
8–12 months
From
₹35,000
Next cohort
Two intakes a year
Explore programme

1:1 Deeptech Mentorship

Built for students who want to build

Direction, not content — a weekly hour with someone who has done it.

Format
One-to-one
Duration
3–6 months
From
₹12,000
Next cohort
Rolling · limited seats
Explore programme

Give a teacher back their week.
Watch what they do with it.

The hours GunanQ removes are not saved — they are redeployed into doubt sessions, remedial classes and research. And the same conviction runs through everything else we do.

0%
Less time on the mechanical part of grading
0%
Of students receive written, personal feedback
0
Best GATE CS rank from our cohorts
0
Grades published without faculty sign-off
Writing

What we've been
working out.

Practical writing on assessment, GATE preparation, research and robotics — from the people doing the work.

All articles
400 APPLICATIONS certificates, mostly THE ONE THAT GETS READ /sim-to-real-arm it runs, and there is a video eleven months of evenings
Careers
9 min read

AI and robotics internships in India — how to actually get one

What research and deeptech teams look for in intern applications, why most get rejected in thirty seconds, and what to build instead of another certificate.

AI-Shala Team Read
IN SIMULATION true state · clean torque · fixed timestep THE GAP ON HARDWARE sensor noise · latency · loop jitter
Robotics Pillar guide
10 min read

Sim-to-real transfer — why your policy fails on hardware

A policy that is perfect in simulation and useless in the lab is not a bug in your training. It is a specific, diagnosable gap. Here is how to find and close it.

AI-Shala Team Read
fifty papers read Submitted Under review Revise Accepted
Research Pillar guide
11 min read

Publishing your first AI paper as an undergraduate

What actually stands between a capable student and a first publication — and the order of operations that gets you from reading papers to submitting one.

AI-Shala Team Read

Work with us. 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.