About us

We’re a research company teaching AI how engineers think.

AI can write code and pass the bar exam, but it can’t design a gearbox. It doesn’t know why an experienced engineer makes the call they make — and that judgment isn’t written down anywhere on the internet. It lives with the engineers who do the work. Teaching it to machines is one of the hardest open problems in AI, and it can’t be solved by ML engineers alone. It takes mechanical engineers.

We’re a Y Combinator company. We work with many of the world’s foundational AI labs. We closed a seed round led by Race Capital, early backers of Databricks and Groq, with angels from OpenAI, NVIDIA, xAI, and Google DeepMind.

https://race.capital/blog/our-investment-in-vision-lab-the-industrial-data-layer-for-physical-ai The role

We’re opening an engineering research group in India, and you are the first person in it.

The problem you’ll work on: define what good engineering judgment actually looks like — precisely enough that a machine can be measured against it — and then build the system that does the measuring. It has never been written down. You’ll write it, test it against real engineering work, and turn it into something that runs without you.

Most design roles use AI as a tool. This one is different: your engineering judgment becomes the standard that frontier AI is trained and measured against.

- Define what good means. Nobody has standardised this. You write the rules, and make them tight enough that two engineers reading the same work land on the same verdict
- Be the bar. Test the standard against real engineering work, every day. Nothing clears until you say it does — and the engineer who produced it gets your reasons
- Build the tooling. Turn what you catch by eye into automated checks, designed with our engineers.



You’re not handed the tool; you decide what it should look for
- Bring in the engineers. India has one of the deepest pools of engineering capability anywhere — design offices, product companies, hardware startups, manufacturers, across Bengaluru, Pune, Chennai, NCR and the belts around them. You bring the good ones into the programme and teach them the standard
- Grow into owning it. The standard, the tooling, the network of contributing engineers, and a team of your own

What we want We’re not screening on years. We want someone early in their career and already exceptional — someone who understands design at the feature-tree level, not the slide-deck level. You don’t need an ML background; your engineering depth is the qualification. Curiosity about AI is what we ask for.

- You’ve designed real products that got made — parts taken from requirement to production, with the revisions in between owned by you. Three years of the right work beats ten of the wrong kind
- Deep in native CAD — SolidWorks, NX, CATIA or Creo: feature trees, assemblies, GD&T;, drawings a factory can build from. And you know why a design changed — draft and wall thickness, mould flow, tolerance stack-up, cost
- You can read a simulation report and say whether it supports the claim — load cases, boundary conditions, mesh convergence
- You can defend a verdict. Telling a senior engineer their work doesn’t clear the bar, and holding that line with reasons, is a large part of this job




- Discreet with other people’s IP: you know what is yours to carry between employers and what isn’t, and you act on it without being told
- You move between worlds easily — part of this job is translating between an Indian design office and a research team in San Francisco, often on the same day. Time studying or working outside India is a strong plus for exactly that reason
- Academically robust — a leading engineering programme in India (IIT, NIT, BITS or equivalent) or a good international university. We hire on ceiling rather than brand, but it is one of the signals we read
- You know this world and people take your calls — design offices, factories, tooling shops, engineers who are actually good. Depth but no network yet? Tell us anyway — we’ll build it with you

Bachelor’s or Master’s in Mechanical, Production, Manufacturing or a related engineering discipline. Helpful, not required: PLM/PDM, test and validation experience, tooling or production exposure, an engineering audience you’ve built.

Comp and logistics

- Salary benchmarked against product companies, not services firms, and set to the person rather than the title. We'll share the band on our first call. Equity for the right candidate
- Based in India, with travel to engineering offices and factories, and trips to our SF HQ
- Work directly with the founders, with exposure to the AI labs defining this field

To apply Send your CV and a few lines on why this role to [email protected]

Not ready for a full-time move? We’re also glad to hear from part-time contributors — engineers who want to review work a few hours a week, and people who can introduce us to engineering teams they rate. Write to the same address and tell us what you’d want to take on.

📌 Mechanical Engineer, Physical AI (Bengaluru)
🏢 Vision Lab
📍 Bengaluru