Selected Work
Hired to build. Shipped to last.
Fractional CTO engagements and products where I designed the interface, built the front and back end, and shipped the AI underneath — for regulated and mission-driven teams that couldn't afford to get it wrong.
Fractional CTO
Product Design + UI/UX
Full-Stack Engineering
AI & Automation
Client names are withheld under NDA. These are my own diagrams, architecture, and results — not proprietary screens.
01
Fractional CTO · AI Build · Client
AI matching that opened two new revenue lines
A nonprofit talent-matching platform needed candidate review and matching to scale past what its team could do by hand — without losing the judgment that made placements good.
Problem
The mission ran on matching talent to opportunity, but review was manual and capped by team capacity. Growth meant either reviewers the budget couldn't cover, or software that scaled the work while preserving human judgment.
What I did
Built an AI review-and-matching engine on TensorFlow and LLMs with the org's own criteria encoded, kept humans in the loop where judgment mattered, and designed it so new services could be layered on the same core.
TensorFlowLLMsPythonMatching / rankingHuman-in-the-loop
Matching engine · dashboard
How it works
Candidates→AI review + score→Human review→Match
2
New revenue streams opened
Read case study →
02
Fractional CTO · Architecture · Client
One codebase, open-source and a business at once
An organization wanted to build in the open and build a company at the same time. I refactored the monolith so community-facing code could live in public repos while trade-secret pieces stayed private.
Problem
One tangled codebase was trying to be an open-source project and a business with protectable IP. They couldn't open the community-facing parts without exposing trade secrets — and couldn't move fast on either.
What I did
Refactored the monolith into services split along a public/private boundary, set up packaging and CI/CD so open-source components could be consumed by the private app without secrets ever crossing over, and established a safe contribution model.
MicroservicesOpen-source architectureCI/CDSecrets managementRBAC
Repo architecture · public / private split
The split
Public repos · OSS↔ published packages ↔Private repos · trade secret
Read case study →
03
Founding CTO · Product Design · Full-Stack · AI
Inkra — conversation to finished work
Co-founder & CTO. An AI platform that turns real conversations — on Zoom, over the phone, in person — into completed, structured work for the non-technical teams running nonprofits.
Problem
Program staff spend hours turning a single conversation into notes, documents, and follow-up tasks. The people doing this work aren't technical and don't want a tool that feels like one.
What I did
Owned it end-to-end: designed the conversational UI and user flows, then built the front end, back end, and the agent layer — with HIPAA-shaped workflows, an audit trail, and a privacy layer built in from day one.
Next.js / ReactFastAPI · PythonPostgreSQLAWS ECS FargateHIPAA / SOC 2
Inkra · product screen
How it works
Conversation→Transcribe→LLM agents→Structured output
Read case study →
04
Product · Full-Stack · AI
Product-validation tools, inside your own AI
An MCP server that exposes a full product-validation pipeline as composable AI tools — usable directly inside Claude, ChatGPT, and Cursor, with no new app to learn.
Problem
Non-technical founders can't self-serve product validation; the knowledge is fragmented and per-user tooling is expensive to run.
What I did
Designed and built the server and its tool set, then distributed it through MCP registries so people use it inside the AI assistant they already pay for — self-serve, with zero per-user infrastructure cost.
MCP protocolComposable toolsRegistry distributionClaude · ChatGPT · Cursor
Tool suite · registry listing
The pipeline, as tools
Idea validationRevenue-velocity scoringDemand testingGTM planningSpec generationTicket creation
05
Product Design · Mobile · Full-Stack
Brand system to app store, in one pair of hands
Cross-platform mobile apps taken from the first brand decision all the way to a store-ready build — design and engineering by the same person.
Problem
Small teams need a polished, on-brand mobile presence and don't have a separate designer, iOS dev, Android dev, and release engineer to make it happen.
What I did
Built the brand system and UI, then shipped the app on both platforms from one React Native codebase — through TestFlight and store-ready EAS builds.
React Native + ExpoiOS & AndroidBrand / UI systemTestFlight · EAS
App screens · iOS & Android
Also built
Work from inside regulated engineering orgs, shown at a high level. Screens stay behind the NDA; the systems and outcomes don't.
Multi-agent platform
agentlib
Six specialized AI agents and seven plugins spanning the full dev lifecycle at a public medical-device company. Adopted as a standard; compressed a 2.5-month security project to six weeks.
Clinical AI
RAG progress notes
Retrieval pipeline with NLP compliance scoring and human-in-the-loop review gates, built to cut provider documentation time in a HIPAA setting.
Community tooling
AI skill-dev bot
A Slack-based grading engine for a 10,000+ member community — automated, personalized feedback across text, image, and voice submissions.
Have a validated idea that needs a builder?
That's the whole job. Let's figure out if I'm the right fit for what you're building.
Book a 30-min discovery call
