~/portfolio/projects-overview.md READONLY

Projects

Case studies first, then supporting public work. The emphasis is on systems that are private, inspectable, and useful outside a controlled demo.

The common thread is pragmatic engineering: preserve context, reduce repeated work, expose tradeoffs, and leave the team with something they can actually maintain.

~/projects/reviewer-notes.json 3_SIGNALS

How To Read This Page

Each project is here for a reason. The goal is not to show volume. It is to show what kind of technical judgment I bring to different kinds of work.

Private AI Work

Project S shows how I handle evidence quality, retrieval design, and system boundaries when the work cannot be treated like a public demo.

Product Restraint

LeXpand shows a smaller but important skill: keep the useful part, reject the bloat, and make trust legible in the product shape.

Operational Packaging

The public infrastructure repo shows how I package repeatable setup work so another operator can actually use it without guesswork.

~/projects/project-s.json FEATURED

Project S (Featured)

Private document-intelligence pipeline built to retrieve from difficult enterprise material, preserve evidence, and improve answer quality before generation starts.

The case study focuses on the parts that usually decide whether RAG is useful in practice: ingestion discipline, hybrid retrieval, reranking, privacy boundaries, and inspectable outputs.

Why it matters: This is the clearest signal of how I approach serious private AI work when trust and maintainability matter more than novelty.

Read Case Study →

~/projects/lexpand.json STABLE

LeXpand

Local-first text expansion workflow focused on privacy, speed, and keeping repeated writing tasks under the user's control.

The case study is less about novelty and more about product restraint: solve a repetitive problem cleanly, stay fast, and keep trust high by defaulting to local control.

Why it matters: It shows product judgment, UX framing, and the ability to choose the smallest useful solution instead of building a noisy platform.

Read Case Study → View Landing Page →

~/projects/supporting-public-work.cfg OPEN

Supporting Public Work

Not every useful signal needs to be a full case study. This repo is here because it shows the same habits in a smaller public package: repeatable setup, local ownership, and clear operating boundaries.

~/projects/cloudflare-tunnels.cfg GITHUB

Cloudflare Tunnel Setup

Guide-driven infrastructure project for locally managed Cloudflare Tunnel configuration and simpler RDP or SSH setup workflows.

A practical utility focused on repeatable setup, local ownership, and making secure remote-access plumbing less error-prone for the next operator.

Why it matters: Smaller scope, same habit: clear boundaries, reusable setup, and work packaged for the next person instead of only the original builder.

GitHub Repository →

~/projects/elsewhere.url OPEN

What To Look For

The strongest signal in these projects is not novelty for its own sake. It is the way the work stays grounded: clear constraints, understandable tradeoffs, and systems shaped around what a team will need once the first version exists.

If you are evaluating fit quickly, read Project S for private AI judgment, LeXpand for product restraint, and the infrastructure repo for packaging and operator empathy.

See Experience Start a Conversation