The build: Our client is building the intelligence layer for construction — model-garden approach, newest computer vision and LLM systems, shipped into prototypes users actually rely on. They're an early-stage SF team, and they want a founding engineer to build it with them.

This isn't a seat. It's a direction. As the founding engineer, you'll set the technical direction, pick the models, and define how the intelligence layer gets built. Expect ambiguity, ownership, and schlep — and the equity to match.

What you'll do

• Own the intelligence layer end to end — detection, understanding, and estimation over construction documents.
  
• Set the technical direction for how the team applies CV, LLM, and AI systems. Model-garden it: pick, test, combine.
  
• Ship 0-to-1 prototypes fast, then harden them from real user feedback.
  
• Coordinate models — detection, segmentation, OCR, LLM reasoning — into one coherent system.
  
• Do the schlep. Early product = messy data, weird edge cases, no clean pipeline, no playbook.
  

Requirements

What we're looking for

• Applied CV/ML in Python — deep learning, computer vision, object detection.
  
• 0-to-1 builder. You've started something, or joined at the start and owned it.
  
• AI in production — you've shipped, not just trained.
  
• Slope > y-intercept. Fast learner beats a static résumé.
  
• Empathy + honesty about what's working and what isn't.
  
• Comfortable with schlep and ambiguity — there's no one to hand it to.
  

Bonus

• You built something people used or paid for (founder/startup).