Founding Software Engineer

The Role

We're looking for a founding software engineer who is obsessed with building agentic AI systems and wants to ship real product from day one. You'll work across the stack but spend most of your time on backend systems, building and orchestrating the AI agents behind Wren, our borrower assistant. Wren talks with borrowers by chat and voice in English and Spanish, checks documents at upload, and gives loan officers and processors a live view of every file, all in a regulated industry.

This is not a "learn on the job quietly" role. We are onboarding lenders and brokers in small groups and building alongside them, so you'll be in the room with customers, see how your code impacts their workflow, and iterate fast. This role is on-site in our San Francisco, CA office.

What You'll Work On

1. Designing and building agentic AI pipelines: tool use, multi-step orchestration, evaluation, and reliability at scale, across borrower intake, document verification, and internal workflows for loan teams
2. Backend services in Python (FastAPI) on Aurora PostgreSQL and AWS
3. Multi-tenant architecture with strong security and compliance boundaries: role-based access, audit logging, encryption, and clean handoffs to licensed staff
4. Integrations with mortgage industry platforms (LOS and point-of-sale systems, voice, messaging)
5. Frontend experiences in React, Vite, and TanStack when the product calls for it, from the borrower application to the loan officer pipeline view

What We're Looking For

1. Strong fundamentals in backend engineering (Python preferred)
2. Hands-on experience building with LLMs: agent loops, tool calling, prompt engineering, evals
3. Comfort with PostgreSQL, schema design, and migrations
4. Exposure to multi-tenancy patterns, auth, or compliance-aware systems is a plus
5. Bonus: sharp design sensibility and strong frontend skills (React, Vite, TanStack). We value engineers who care deeply about craft on both sides of the stack.
6. High energy, low ego, genuine curiosity about how things work
7. You care about the people using what you build

About the interview

We run a quick interview process. You can expect 1-2 calls with the founders, 1 technical interview, then 1 work trial onsite in SF.

About Loanwren

Getting a mortgage should not depend on how well a borrower keeps up with an inbox full of document requests. Yet origination still runs on back and forth: a loan officer asks for paystubs, the borrower sends a blurry photo, a processor chases the missing page, and underwriting adds conditions at the end. Borrowers give up, loan officers lose hours to follow-ups, and closings slip.

Loanwren is the AI-native platform that guides borrowers through intake, verifies documents as they arrive, and delivers cleaner files to loan officers. Wren answers borrower questions at any hour, asks only for what the file needs, and fills in forms from uploaded documents. Each document is checked on arrival: fields extracted, names and accounts matched, anything unusual flagged for the team. Loan officers and processors see every borrower, every document, and every flag on one screen.

We believe AI should do the repetitive work of collecting, extracting, and checking, while licensed professionals make every lending decision. Loanwren is a software platform, not a lender, broker, or credit decision-maker, and our assistant is built to hand off rather than guess. We are a small team, now onboarding our first lenders and brokers, and we are just getting started.