STAFF ENGINEER · AI / ML
Own the modeling stack behind ROOF10X Estimate — the AI system that turns aerial imagery and homeowner inputs into a sealed, signable roof bid in under 90 seconds.
ABOUT THE ROLE
Estimating is the heart of the product. A homeowner files a claim, our system pulls aerial imagery, runs measurement and damage detection, prices materials and labor against the contractor's catalog, and returns a bid the rep can text from the driveway. Every percentage point of accuracy compounds across thousands of jobs.
We are looking for a Staff-level engineer to own this stack end-to-end: data, models, evaluation, and deployment. You will set the technical direction for how we close the gap between the AI estimate and the as-built invoice — currently around 6%, and the contractors who bid us against legacy software notice the difference immediately.
WHAT YOU’LL OWN
- →Own the measurement, damage-detection, and pricing models that power Estimate. Decide what we build, what we buy, and what we replace.
- →Run a tight evaluation loop: ground-truth dataset, regression suite, shadow-mode comparisons against active jobs.
- →Pair with the data team to instrument every contractor invoice that comes back so the model gets sharper every week.
- →Mentor two product engineers ramping into ML work. Set the bar for code review on inference paths.
- →Deploy on Cloudflare Workers AI plus a self-managed GPU pool for vision; you decide where the line moves.
WHAT WE’RE LOOKING FOR
- →8+ years building production ML systems. At least one role where you owned a model end-to-end (data → training → eval → deploy → monitor).
- →Strong applied vision experience — segmentation, object detection, or multi-view geometry.
- →You write production Python and at least one of Rust, TypeScript, or Go without complaint.
- →Comfortable picking the right tool: classical CV when it works, a small fine-tuned model when it doesn't, an API call when neither is worth the maintenance.
- →Calm in the face of messy real-world data. Roof imagery is gnarly.
NICE TO HAVE
- +Experience with aerial/satellite imagery, photogrammetry, or LIDAR.
- +Worked on a system where model errors translated directly into dollars.
- +Background in construction, insurance, or claims tech.
HOW WE INTERVIEW
- 0130-min intro with the hiring manager.
- 02Take-home: ship a small evaluation harness on a sample dataset we send you. ~4 hours, paid.
- 03Technical loop: systems design, applied ML, and a working session on a real Estimate failure case.
- 04Final: founder conversation on scope and ownership.