Projects
Anchoring AI
A build-with-me series documenting a RAG pipeline and MCP server, built as a portfolio mirror of consulting work I did for a credit union. Four posts so far, covering the architecture, embeddings and retrieval, the MCP interface layer, and running models locally. More in progress.
View project →Newsroom
A full AI content pipeline that carries a rough idea to publishable output through a deterministic six-stage workflow — oracle, interview, draft, editor council, finalize, lessons — with a human at every gate. Runs on a compare-and-set state machine, a background-job engine with per-run cost ceilings, and swappable multi-provider model routing, over a Git-native brain. Next.js + Python/FastAPI, MongoDB, Google Docs/Drive. Built solo alongside an AI agent crew.
View project →Masthead
A git-native "brain" for an AI writing system: the machine's whole mind — interviewer and editor personas, voice packs, engine steps, in-flight drafts, and safety rules — lives as plain Markdown whose prose is read verbatim as model prompts, with git history as the memory. This is the neutral public demo instance, showing the architecture end to end.
View project →Credit Union MCP Server
A shared enterprise MCP server, in development, that will expose call recordings, transcripts, and CSAT data to a credit union's internal GenAI teams through one common interface. Auth via Microsoft Entra ID; targeting the client's Azure subscription.
AWS Marketplace Launch - GPU Analytics Database
Built and published the AWS Marketplace listing for a GPU-accelerated analytics database: AMI construction, cloud-init scripting for instance lifecycle events, non-pay AMI variants for POC use.