Founding AI Engineer

Somewhere in America right now, an apartment building is waiting.

Not on concrete. Not on financing. On a human being reading a stack of blueprints line by line, checking them against a code book, and eventually writing back with corrections. That round trip takes close to a year.

Our client is building multimodal AI agents that read architectural drawing sets and check them for code compliance — turning a months-long review into a single day. If it works, housing gets built faster. Not as a mission-statement abstraction. As buildings.

The team: 3 people. An oversubscribed $4.3M seed led by a respected urban-innovation fund, angels behind category-defining mapping and infrastructure products, and founders who are pro-housing advocates with a real record of changing housing policy in California. They didn't discover this problem in a market map — they hit it while trying to get housing built.

You'd be one of the first 1–3 engineering hires.

What you'll actually do

Every week is a loop: form a hypothesis about why the agent misread a drawing, run a time-boxed experiment, measure it honestly, and ship what wins.

• Improve accuracy of multimodal agents that read and reason over architectural documents
• Work across supervised fine-tuning, distillation, auto-research, retrieval, and context engineering
• Run computer-vision evals and build the internal tooling that makes everyone else's experiments faster
• Direct your own research roadmap — if an experiment doesn't duplicate work or block someone, nobody says no
• Talk to users and dogfood the product weekly

Your work ships to real users inside your first month. There is no six-week onboarding.

You'll fit if

• 2–8 years in software engineering or ML, focused on AI/LLM applications
• BS in CS, ML, or a related technical field — or equivalent practical experience
• Production LLM / AI-agent experience (stack: Python, TypeScript, Next.js, LangGraph)
• You use Claude Code, Cursor, or similar every day — here they're infrastructure, not a novelty
• You've designed and run ML evals and experiments
• You blend SWE speed with ML rigor: not too researchy, but never skipping the measurement
• Highly autonomous — you generate your own experiments and manage your own workload

Why take this seat

• A real founding stake. 0.75–2% equity at a $30M cap, on a three-person team where the founders still own most of the company.
• Physical-world impact. Your model's accuracy is measured in how fast actual apartments clear review.
• Maximum autonomy. Your own research roadmap from week one.
• Taken care of: $165K–$210K base, fully paid health insurance with $1,000+/yr HSA contribution, free lunch and dinner, 401(k), commuter benefits, monthly wellness benefit.