Securities Regulator → AI Architect
Governed AI orchestration for regulated finance
I design auditable pipelines where models interpret and deterministic code executes. The judgment of a former securities regulator, applied to how AI actually ships inside finance.
Selected work
A natural-language-to-SQL agent where the model only classifies intent and drafts candidate queries; deterministic Python and SQL handle execution, validation, and logging — every step auditable.
A 20-part interactive series teaching governed AI orchestration for banking and payments analytics — built from a typed-data generator, not hand-authored pages.
A governed multi-agent financial-crime research assistant: a supervisor delegating to retrieval, structured-data, and analytics sub-agents, with a validation judge and human-in-the-loop over an auditable trail.
Lab
Gesture and wearable control for wall projection — MediaPipe hand tracking and an IR night-vision camera, with an ESP32 + IMU wristband as the next-gen input.
A WebGL kaleidoscope for psytrance and techno projection, driven live by Wiimote and gesture input.
I spent four years as a capital-markets regulator at Peru’s securities authority — designing IOSCO-aligned rules, authorizing funds, and supervising transaction monitoring — before moving into building AI systems for the same industry I used to oversee.
That combination is the point: deep, current LLM engineering paired with genuine regulated-finance domain authority. I build systems that are powerful because the model interprets, and trustworthy because deterministic code executes, validates, and logs.