Founding AI Engineer - Staff

About the Role

The engineering problem: general-purpose LLMs aren't good enough for legal work in their default state. We're building expert-in-the-loop systems which capture lawyer feedback continuously, feed it into our knowledge base so that the right lawyer guidance can be provided to LLMs via an agentic RAG process. The process involves a robust evaluation process to ensure high quality. As a Staff AI Engineer, you'll own the core AI systems that solve this problem. You'll set the technical direction for how we build, evaluate, and ship AI that works alongside lawyers in production.

What You'll Do

1. Build and maintain LLM-based agents that handle complex legal workflows with lawyer oversight at critical decision points
2. Design expert-in-the-loop systems where attorney feedback continuously improves AI quality — the core differentiator of our product
3. Build evaluation infrastructure including LLM-as-a-judge evals for continuous monitoring, internal tooling for collecting feedback, and frameworks for running eval sets efficiently
4. Stay on top of the frontier — track latest LLM architectures and model updates, build a framework to evaluate potential enhancements to the product
5. Work directly with lawyers and product to understand how AI should behave in legal contexts and translate that into engineering decisions

What We're Looking For

Required

1. 5+ years of production backend engineering experience in Python — you write code that ships and scales, not just notebooks
2. 1+ year building LLM-based agents in production - agents that call external APIs, manage state, and make decisions. 
3. Built evaluation systems for measuring prompt and agent performance pre and post deployment.
4. Tech lead experience: you've owned architecture decisions, led development, handled deployment and monitoring
5. AI-native workflow: you're already writing 90%+ of your code with tools like Claude Code, Cursor, or Codex
6. In-person in SF: we build together in the office — this isn't a remote role

Nice to Have

1. Experience building planning-based agents — systems that reason about goals, decompose tasks, and decide on actions autonomously
2. Experience designing systems with expert/human feedback loops — where domain experts continuously improve AI output