Founding Engineer

Agentic document systems. We're building LLM-powered pipelines that ingest, classify, extract, and generate complex financial documents — subscription agreements, capital call notices, K-1s. The challenge isn't getting a demo working. It's getting 99%+ accuracy on documents where a single wrong field is a compliance violation, across hundreds of fund-specific variations, with a human-in-the-loop review system that's fast enough to be useful.

Automation that replaces manual operations. Our clients' workflows are deeply manual. We're systematically identifying high-volume, high-cost operational processes and building systems that eliminate them — not by shipping a chatbot, but by building structured automation with robust error handling, audit trails, and graceful fallbacks. Every automation we ship directly reduces burn and increases margin.

Platform modernization under load. We inherited a large microservices codebase. We're migrating toward a modern, agent-friendly architecture while keeping production stable for clients processing millions in transactions. This is systems engineering, not greenfield hacking — and it requires judgment about what to rebuild, what to wrap, and what to leave alone.