Staff Engineer, Agentic Intelligence - San Francisco

Location: San Francisco - hybrid, in-office two days a week

Compensation: Competitive base + leadership-level equity

Mortgage underwriting is the work of deciding whether a loan can be approved - checking a borrower’s income, assets, credit, and collateral against a dense web of rules and documents. We’re teaching an AI agent to do that work. But an agent is only as good as the policies that govern it. The hard part isn’t getting a model to produce an answer - it’s encoding underwriting judgment into the policies the agent reasons against, and building a framework flexible enough that those policies can be configured, customized, and changed as fast as the business and its lenders demand.

That’s the problem we’re solving.

HomeVision is the AI underwriting platform for the $13 trillion U.S. mortgage industry. Our platform, MIRA, has already automated more than two million appraisal reviews for many of the country’s largest lenders and appraisal management companies, reducing review time by 75% and doubling underwriter productivity.

Now we’re expanding beyond appraisal review to cover the full underwriting workflow - income, assets, credit, collateral, and loan file decisioning - to create the first end-to-end AI underwriting platform for the industry. MIRA’s been adopted by dozens of real enterprise customers with contracted production demand, including Newrez, the fourth-largest mortgage lender in the U.S. and a strategic investor in HomeVision, alongside our venture backers, Initialized Capital and NVP Capital.

We’re looking for a staff-level engineer to own the intelligence and policy layer of our underwriting agent - the policies that encode underwriting judgment and the configurable framework that lets us author, customize, and dynamically change them across lenders, investors, and loan products. You’ll partner directly with our ML engineers, our CTO, and senior product leadership to define and execute our agentic roadmap.

• Own the policy framework. Design and build the framework that turns underwriting judgment into policies the agent reasons against - structured so policies can be configured per lender, investor, and loan product, customized for edge cases, and changed dynamically without a code deploy.

• Author the policies themselves. Translate dense, real-world underwriting rules - including escalation paths and human-in-the-loop boundaries - into clear, testable policy that the agent executes reliably. The difference between an agent that’s clever and one that’s trustworthy in a high-stakes financial workflow.

• Make policy quality measurable. Establish and tune evals that prove a given policy behaves correctly across messy, real loan files. Build the datasets and harnesses that measure it, and turn results into a fast iteration loop.

• Own the agent harness it runs in. Maintain the runtime the agent operates in - tool use, orchestration, context management, retries, and guardrails - so policies execute reliably and safely on complex, multi-step tasks.

• Elevate the team. Establish the patterns, libraries, and review standards the rest of the team builds agents and policies against. Mentor and recruit the engineers who’ll work alongside you.

• Ship fast and learn faster. Take capabilities from rough ideas to production in days, not weeks. Watch how they perform on real files and rapidly iterate.

• You’re deeply product-oriented. You enjoy engaging with customers and learning a new domain. You think beyond implementation details and care how technical decisions shape outcomes and trust.

• You’re a builder at heart. You’ve shipped meaningful systems used by real customers. You likely have side projects, strong opinions about technology, and genuine curiosity about where AI is heading. You are a creator.

• You operate with urgency and ownership. You move quickly without sacrificing reliability or trust. You proactively identify problems, communicate clearly, and drive solutions.

• You elevate the people around you. You bring strong engineering judgment, high standards, and collaborative energy. Teams become stronger and more engaged when you’re involved.

• You’re excited by difficult workflow problems. Agent reliability, evaluation under ambiguity, encoding judgment as configurable policy, and AI-assisted decision-making genuinely interest you.

• Strong engineering proficiency. Your coding skills are top-notch, as is your ability to wield AI-powered coding tools to expand your impact.

• Experience building and operating LLM-based or agentic systems in production.

• Experience designing configuration-driven systems - rules engines, policy/DSL frameworks, feature-flag or config platforms, or similar systems where behavior is driven by data, not redeploys.

• A rigorous, empirical approach to evaluation - you know how to measure quality in systems where “correct” is hard to define, and you trust data over vibes.

• Comfort working across the stack when it matters - you’re comfortable moving between our Go/Python backend and our TypeScript frontend to build holistic solutions.

• Track record of leading a technical surface end-to-end - not just features. Bonus if you’ve worked on decisioning, document understanding, or human-in-the-loop products.

We’re a hybrid team headquartered in San Francisco, with team concentrations in Seattle and Argentina, with other great team members in parts of the U.S.

Our values are the ones we actually use:

• Start with questions, not answers.

• Run at problems - make your footprint bigger than your foot.

• Continuous learning, continuous improvement.

• Be transparent and collaborative.

• Value time.

• Play to win, win for everyone.

• Always do the loving thing.

We’re growing the engineering team meaningfully over the next twelve months and this role will have a direct hand in shaping who that team becomes.

• Competitive compensation calibrated to senior/principal-level engineers in San Francisco

• Meaningful equity ownership

• Full medical, dental, and vision coverage

• 401(k) with company match

• Flexible vacation policy

• Parental leave

• Learning and professional development support

Send a note, a résumé, and (ideally) links to things you’ve built. GitHub, side projects, products you’ve shipped - whatever lets us see how you think and build.

We’re thoughtful and deliberate about who joins the team, but we move quickly when we find the right person.