Mason AI — Founding AI Engineer

davidjoseph-co; San Francisco (CA); On-site; USD 160,000 - 200,000; Full time

Benefits offered by this job

Free lunch and dinner

Health insurance with HSA

Commuter benefits

Wellness stipend

Job summary

Mason AI in San Francisco, CA is seeking a founding AI engineer to own the full experiment loop for improving multimodal blueprint‑compliance agents. This in-person role requires deep ownership, rapid experimentation, and strong evaluation skills to drive progress in a tiny, mission‑driven startup.

Compensation ranges from $160,000 to $200,000 with 0.5–2% equity, plus full benefits and relocation support. The position demands ~55 hrs/week, daily on-site presence, and a passion for solving the

Qualifications

• Independent and self-directed founding engineer capable of driving experiments end-to-end.
• High velocity prototyping with rapid iteration cycles and low supervision.
• Rigorous evaluation mindset to steer high-risk, high-reward work.

Responsibilities

• Own the full experiment loop to improve Mason's multimodal blueprint‑compliance agents.
• Develop autoresearch, run sweeps, and analyze results to improve accuracy.
• Build domain knowledge of ADA, Wildland-Urban Interfaces, and building codes.
• Wear multiple hats: CV experiments, lightweight data engineering, tooling, and UI work.
• Engage in weekly user contact and product dogfooding as needed.

Skills

Independent and self-directed

Fast prototyping

Rigorous evaluation

TypeScript familiarity

In-person SF work

Tools

Claude Code

Eval framework

Web tooling

Job description

Mason AI - Founding AI Engineer

Type: Full-time | On-site | San Francisco, CACompensation: $160,000-$200,000 + 0.5-2% equityHiring count: 1Visa sponsorship: None available. Relocation supported (per outreach template).Reports to: Not specified on role page - founding role; team of 3 moving to 2 as the CTO transitions out.

About Mason AI

Mason AI is building agentic AI for the built world - multimodal agents that review construction blueprints for buildability and code compliance, becoming domain experts on rules like the ADA and Wildland-Urban Interface. The company grew out of SFYIMBY, the pro-housing organization in San Francisco, and its mission is to solve the housing crisis: if Mason wins, the cost of building housing drops and rents get more affordable. It is the technical leader in its space with the most progress of any competitor.

Founded: 2025 | Team size: Seed 1-10 (currently 3, moving to 2) | Total funding: $4.3MIndustry: Property TechWebsite: withmason.aiOffice: San Francisco, CA

Why Candidates Should Join

• Work directly on housing: One of the very few places in tech where an engineer works directly on the housing crisis - a mission shared even by its investors, and the company's single strongest draw.
• Founding ownership: A large chunk of equity (0.5-2%) and a founding-engineer mandate; the company is explicitly betting its future on this hire and the next.
• Technical leadership + strong demo: Furthest-along player in its space, with a product demo that reliably gets strong engineers excited.
• Comp + full benefits: $160-200K base plus free lunch and dinner at the office, fully paid health insurance with $1,000+/yr employer HSA contributions, commuter ($100/mo) and wellness ($50/mo) benefits.

Intake Call Summary

• Deeply self-directed founding engineering role at a tiny company: currently 3 people, moving to 2 as the CTO transitions out.
• Core of the job is running experiments end to end on Mason's eval framework - generating hypotheses, implementing them, and rigorously analyzing results to improve multimodal blueprint-compliance agent accuracy.
• Two profiles have worked here: (1) an exceptional SWE who wants to break into AI research, and (2) an AI/ML engineer already established in the space.
• Heads-down and technical: only ~2 hrs/week talking to users (occasionally on-site) - this is not a forward-deployed role.
• High intensity: ~55 hrs/week with weekend availability, in person in SF every day, in exchange for large ownership.
• The company is explicitly betting its future on this hire and the next.

The Role

A deeply self-directed founding AI engineering role owning the full experiment loop to improve the accuracy of Mason's multimodal blueprint-compliance agents.

What You'll Be Doing

• Spend the majority of your time improving the accuracy of Mason's multimodal blueprint-compliance agents - building autoresearch, improving evals, and running sweeps on experimental features you design.
• Own the full experiment loop: coming up with ideas, implementing them, and rigorously analyzing results on the eval framework.
• Become a domain expert on the parts of the building code that matter (ADA, Wildland-Urban Interfaces, structural and fire safety).
• Wear hats as needed: computer-vision experiments, lightweight data engineering, internal tooling, and webapp improvements.
• Weekly user contact and product dogfooding, roughly two hours a week.

Tech stack: Tech-stack agnostic; TypeScript front-end and back-end referenced. AI coding tools (e.g. Claude Code) used in the process.

Requirements

• Independent and self-unblocking, able to direct your own work with minimal oversight
• High velocity, fast at prototyping and shipping experiments
• Rigorous about evaluation, with genuine research taste to steer high-risk, high-reward work
• Either a strong SWE eager to grow into ML/AI, or an ML/AI engineer already fast at prototyping
• Tech-stack agnostic (comfortable across TypeScript front-end and back-end)
• Genuine mission alignment with solving the housing crisis
• Able to work in person in San Francisco, ~55 hours/week with weekend availability

Green Flags

• In-network candidates and warm referrals (historically the strongest source of hires)
• Deep, authentic mission alignment on housing, which has let Mason land candidates it would otherwise have no shot at
• Self-driving car company backgrounds (Waymo, Zoox): fast to pick up the frameworks and instinctively understand why evaluation rigor matters
• Highly regulated application-layer AI (medical, legal) and climate-tech backgrounds
• Teams where everyone has worked on successful AI products before
• Strong schooling
• The curiosity of a policy wonk to digest building codes and architectural diagrams

Red Flags

• Mercenaries: candidates whose motivation is simply breaking into the startup scene rather than the mission. Some have passed the coding challenge and still been turned down for this reason
• Big-tech backgrounds without genuine appetite for startup pace; the company has been repeatedly disappointed here
• Anyone who needs structure and direction rather than unblocking themselves
• Candidates who cannot or will not commit to in-person SF work at startup intensity

Role Details

• Salary - $160,000-$200,000
• Equity - 0.5-2%
• Experience - 2-8 years
• On-site policy - In person in SF every day, ~55 hrs/week with weekend availability
• Visa sponsorship - None available; relocation supported
• Employment type - Full-time
• Location - San Francisco, CA

Screening Questions

None provided on the Contrario role page.

Interview Process

Stage 1 - Pending Approval - Candidates awaiting initial approval.Stage 2 - Intro Call (30 min) - Informal chat on interests and experience.Stage 3 - AI Coding & Culture Interviews (90 min total, back to back) - Build a prototype for a slice of Mason's product using AI coding tools (e.g. Claude Code), sharing your screen, plus a conversation on how you've navigated challenge.Stage 4 - Paid Work Trial (3 days) - Run evals for the multimodal agent in Mason's repo, generate experiment hypotheses, run them, and analyze results; evaluated on the speed of running thoughtful experiments. Flexible scheduling, often run over a weekend.Stage 5 - Offer ExtendedStage 6 - Candidate Hired - Candidate accepts and starts.

Ideal Companies & Backgrounds

No Ideal Companies section was present on the role page. Background signals drawn from Green Flags / Nice-to-Haves:Self-driving / AV - Waymo, ZooxRegulated application-layer AI - medical, legalClimate tech - mission-oriented startups