01 / SELECTED WORK

The thinking behind the build.

Applied AI across policy, financial investigation, and maritime intelligence. Explore the problem, the engineering, and what the evidence supports.

01 / Agentic AILocal live-model testing

LedgerLens

Observable AI Financial Assistant

A React/TypeScript interface, Node.js/TypeScript Claude agent, and Python retrieval service over DuckDB and ChromaDB.

Locally exercised with live model calls; an operating AWS deployment is not verified.

ReactTypeScriptNode.jsPython
02 / Policy AIIn development · private implementation

PolicyMesh AI

Agentic Policy Analysis & Assurance

Independent research into policy analysis, evidence provenance, explicit assumptions, validation, and human oversight.

In development: independent research project with private implementation.

03 / Financial AIResearch project

AegisAgent

Fraud Investigation & Evidence Review

A Python/FastAPI and DuckDB/dbt research application combining fraud-model evaluation, Bedrock narrative drafting, and human-review controls.

Locally runnable portfolio/research project; current documentation says it is not publicly deployed.

PythonStreamlitFastAPIDuckDB

ADDITIONAL WORK

ALSO BUILT

This digital twin.

The site you are using is itself an applied AI project: a public representative with saved conversations, grounded background information, and controls to export or delete your chat.

Next.jsCloudflareAWS LambdaBedrockDynamoDB

Source code private. This public website demonstrates the application; the implementation repository is not publicly accessible.

Ask my twin