Send something real
Email with whatever best shows your work — repo, paper, thesis, a video of a thing you made move. A CV alone tells us less than one project you can talk about properly.
Come find out what actually breaks.
We're a small team in New Delhi building robots that have to work in real rooms. The work is unglamorous more often than not, and the feedback loop is a machine either doing the thing or not doing it.
Short loops, real hardware, no theatre.
You will spend real time on the floor. Policies get tested on a machine that can drop things, hit things and wear out — which means results are unambiguous and progress is honest.
We're small enough that scope isn't handed down. If you find the failure mode, you usually own the fix. That cuts both ways: there's no one to hide behind, and no one blocking you either.
We publish. Work here has gone to Humanoids and is headed for ICRA, and we'd rather people build a public record than sit on it.
Five ways in.
All roles are on-site in New Delhi. If none of these fit but you think you should be here, the last one is for you.
Robot Learning Engineer
Imitation and reinforcement learning for manipulation. You'll own policies from data collection through sim training to the moment they run on real hardware — including the part where they don't.
Perception Engineer
Multimodal sensing and state estimation. Fusing RGB-D, tactile and proprioceptive streams into something a controller can trust when the lighting is bad and the scene has moved.
Controls Engineer
Whole-body control and compliant actuation. Locomotion and manipulation that stay stable under contact, and degrade safely rather than dramatically.
Mechatronics Engineer
End-effectors, sensor mounting and the integration work that decides whether a good policy has a body worth running on. Comfortable with CAD, wiring and a bench.
Open application
Tell us what you've built and what you want to work on. Interns and new graduates included — we care more about what you've actually made move than where you studied.
Worth being honest about.
Robotics attracts people for the wrong reasons as often as the right ones. Read this before you write.
- You'd rather debug a real failure on hardware than tune a benchmark.
- You're comfortable owning something end-to-end before it's fully specified.
- You think reading papers and shipping code are the same job, not two.
- You want your name on published work and don't mind it being scrutinised.
- Being early and slightly under-resourced sounds like leverage, not risk.
- You want a defined ladder and a stable brief — neither exists yet.
- You'd like to work remotely; the hardware is here, so we are too.
- Demo videos matter more to you than a system that runs for weeks.
- You expect the tooling to already be good. Often you'll build it.
- You need certainty about what the company looks like in three years.
Four steps, about two weeks.
A conversation, not a quiz
Forty-five minutes on what you've built and what you'd want to build here. You should be interviewing us just as hard.
One technical deep-dive
A real problem from our backlog, discussed together. No take-home that eats your weekend, and no whiteboard algorithm trivia.
Come to the lab
Spend time with the team and the hardware. If the work doesn't excite you in the room, it won't after you join.
Write anyway. We read everything.
One paragraph on what you've built and what you want to work on is plenty to start.