Independent AI practitioner building production-grade agent systems — with the retry logic, evals, and runbooks most demos leave out. Formerly PM at AWS, Chief of Staff at an early stage fintech startup.
Each one ships with the failure modes documented, a quality gate, and an answer to "how do you know it's still working."
Turns a company name into a validated lead brief in seconds — replaces 30–45 minutes of manual SDR research.
Classifies incoming client type, routes to the right checklist, then runs contract, workspace setup, and kickoff drafting in parallel.
A stateful agent that finds grant opportunities, retrieves matching RFPs, and drafts narrative sections — remembering prior applications across sessions.
Four specialized worker agents — news, reviews, pricing, feature diffs — synthesized into one weekly competitive brief with traceable sourcing.
I spent my career in operating roles before I spent it in code — running product at AWS, and previously as Chief of Staff inside a Series A fintech lender. That's the lens I bring to agent systems: the question isn't whether it works once, it's whether it keeps working.
Now I build and operate AI agent systems with the eval loops, retry logic, cost and monitoring layer that make them safe to hand off.