Software engineer · Systems, AI infrastructure & product
I build systems that prove what happened and stop what shouldn't. For years that meant compliance and permissions around money. Now it means guardrails for AI agents.
Nanny is an open-source authorization and audit layer for AI agents. You declare which tools an agent may call and under what rules. If it tries anything else, Nanny stops the run and records why. No soft warnings, no negotiation.
# nanny.toml [tools] allowed = ["web_search", "save_findings"] [tools.web_search] max_calls = 30 reads_untrusted = true
Lead engineer on the admin platform the compliance and operations teams run on. Got the B2B dashboard off the ground, shipped its audit logging, and worked across payouts and deposits infrastructure.
Owned the architecture of a crypto payroll product for global teams, from React to Postgres, and led planning, delivery and mentorship.
Built the staking flows behind a ~15% lift in purchases, and the UI kit used across the product.
Shipped production UI in React and Redux.
Works with TypeScript, Rust, Python, React, Next.js, Astro, Tailwind, Storybook, Node.js, NestJS, Express, FastAPI, GraphQL, PostgreSQL, MongoDB, Redis, Prisma, SQLAlchemy, LangGraph, MCP, Vitest, Pytest, Docker, GitHub Actions.
Trained at Lambda School (now Bloom Institute of Technology), full-stack web development and computer science, 2019 – 2020.
Building agents, or something that needs a hard line around what it may do? I'd like to hear about it.
Away from the keyboard: books, poetry, fantasy series, anime, music, and time in nature.