Founding AI Data Engineer - Optexity (India)
Company: Pear VC
Location: Bangalore
Type: Full-time
Posted: 2026-08-12
About this role
## About Optexity
Optexity is a product-driven research lab building clinical reasoning and computer use models on data no other lab can reach. We deploy directly inside clinics and hospitals — including systems with no APIs, using integration infrastructure we've built and open-sourced — and in return become their preferred partner. That gives us proprietary clinical reasoning trajectories from practicing physicians, and a feedback loop between real patient encounters, our models, and the products built on top of them.
We're a small, fast-moving founding team with multiple published papers in NeurIPS, ICML, CVPR etc and background from Apple, Amazon, Microsoft, CMU, IIT.
We are backed by world-class investors and leaders like Jeff Dean, Neotribe VC, PearVC, Together Fund and Zapier Fund.
## How we work
- **Customer obsession** — we start with the customer and work backwards
- **Intellectual honesty** — ideas matter more than titles; we communicate directly and assume good intent, even in disagreement
- **Bias for action** — we build and learn with customers rather than debate in the abstract
- **Extreme ownership** — we own outcomes, not just tasks, and see problems through
Why this role exists
Most research roles at this stage hand you a dataset everyone already has and ask you to be marginally better than the last person who tried. Here you get data nobody else has — real clinical reasoning trajectories from practicing physicians — and the room to figure out what to do with it. This is a founding research hire: you'll define the agenda as much as execute it, with direct founder access, real compute, and nothing between an idea and an experiment.
What you'll do
- Work with real, proprietary clinical data from hospital and clinic partners to surface insights that shape model and product direction
- Build clinical reasoning models that improve physician and clinic workflows
- Design and publish benchmarks that expose where current LLMs fall short on...