Careers
Robots will learn from what we record
The models are ready. The data is not. We are building the network that captures how real work is actually done across India — and handing it to the teams building the next generation of robots.
Why this work
A problem that cannot be solved from a desk
Most AI work happens in a browser tab. This does not. The hardest problems here are on a factory floor at two in the afternoon, in a kitchen during dinner service, in a field during monsoon.
The bottleneck is real
Frontier models are no longer limited by compute or architecture. They are limited by real-world physical data. That is the problem we work on, and almost nobody is doing it well.
Operations is the product
Anyone can buy cameras. Recruiting hundreds of participants, getting site access, running consistent protocols and rejecting bad data — that is the hard part, and it is engineering as much as logistics.
Built here, used everywhere
The data we collect in Pune, Coimbatore and rural Maharashtra ends up training robots deployed worldwide. Very few Indian companies sit that far upstream in the global AI stack.
We say what is true
We tell customers what we do not have yet. We tell participants exactly what their data is used for. If a batch fails QA we re-shoot it rather than quietly shipping it.
Small team, real ownership
You will own a programme end to end rather than a ticket in someone else's backlog. That also means the work is unglamorous some weeks.
The people in the data matter
Every participant is paid fairly, consents in their own language, and can withdraw. That is a constraint we design around, not a box we tick.
Open roles
Where we need people right now
If you are close but not an exact match, apply anyway and tell us which part you would grow into. We read every application ourselves.
Field Operations Lead
Own a collection programme end to end — site access, participant recruitment, scheduling, supervision and daily throughput.
Participant Recruitment Manager
Build and run the participant network — sourcing, screening, consent, training and retention across multiple states and languages.
Site Partnerships Manager
Open doors into factories, warehouses, kitchens, farms and clinics. Negotiate access, agreements and on-site logistics.
Data Quality Analyst
Review episodes against the acceptance spec, decide what ships and what gets re-shot, and feed failure patterns back into protocol design.
Backend Engineer — Data Pipeline
Build the ingest, QA and delivery pipeline: resumable uploads from weak networks, integrity checks, versioned storage, customer delivery.
Computer Vision Engineer
Automate the QA engine — sync drift detection, occlusion and exposure checks, calibration validation, automated redaction at ingest.
Hardware Engineer — Capture Rigs
Design capture hardware that survives a factory floor: multi-sensor sync, calibration that holds, batteries that last a shift, field-repairable.
Full-stack Engineer — Internal Tools
Build the tools operations runs on: scheduling, consent tracking, QA review interfaces, and the customer-facing progress dashboard.
Data Collection Scientist
Turn a customer's one-line task into a runnable protocol — trial structure, sensor placement, edge-case coverage and the acceptance spec.
Robot Learning Researcher
Train policies on our own datasets to find out what is actually missing, and turn that into what we collect next.
Enterprise Account Executive
Own relationships with robotics and frontier AI labs in the US and Europe. Technical enough to scope a protocol on the first call.
Founder's Office
Work directly with the founders on whatever is most broken that month — pricing, hiring, a new sector, a customer escalation.
No open roles in this team right now.
How we hire
Four steps, about two weeks
No take-home that eats your weekend, and no rounds that exist only to test stamina. You will hear back either way.
Apply
Email us with a CV or a link. Tell us which part of the role you would be strongest at and which part you would be learning.
Intro call
Thirty minutes on what we are building and what you want next. Ask us the awkward questions here — funding, runway, what is not working.
Working session
Ninety minutes on a real problem from the job — a protocol to design, a pipeline to debug, a site to scope. Paid if it runs longer.
Team conversation
Meet the people you would work with day to day, including someone from a different team.
References
Two conversations with people who have worked with you. We tell you before we call anyone.
Offer
Written offer with compensation, scope and what success looks like in the first six months.
Nothing above fits?
Tell us what you would build here and why it matters. We have hired people for roles that did not exist until they wrote to us.