Polymr

tl;dr: We're building the AI-native procurement for the factories that run the physical economy but and still run their purchasing on spreadsheets, email attachments, and WhatsApp forwards. Our system reads the documents (vendor quotes, POs, invoices, drawings), reconciles them against each other, and runs the workflow with humans approving the calls that matter.

What you'll own: Document ingestion and extraction , agentic workflows and the eval harnesses that decide when the system acts alone versus asks a human.

• Live, not a demo: our customers run real daily procurement through Polymr. The documents are messy, the edge cases are real, and the feedback loop is measured in days.
• Founder domain-proof: engineers and researchers from Amazon, Microsoft, Goldman Sachs and ServiceNow (Berkeley, Duke, Columbia, Harvey Mudd, Georgia Tech) working alongside supply-chain operators who ran procurement and IT at P&G, Tesla, Otis, and DaVita.
• Early enough to matter: you'd be engineers #3–6. There is no legacy codebase and no committee. What you build is the product.

How we work

Roughly 80% building, 20% embedded with customers, every engineer here talks to the ops managers and purchasing heads who use what we ship, and sometimes sits on the factory floor watching them use it. We use AI coding tools (Claude Code, Cursor) as a daily multiplier and expect you to. [SF/in-person].

Our AI policy

draft your application yourself, polish it however you like; use AI freely on the build round, the way you would on the job; live rounds are just you. PS: we read every application ourselves. Non-AI-written answers stand out!

Three founding roles are open, the titles signal center of gravity, not a lane. Exceptional talent shows up in a lot of forms: shipped products, research, open-source, startups, competitions, or other deeply technical work all count.