Builder · Operator · AI consultant

I build AI systems for a living, and I run what I build.

By day I lead NBA player-tracking deployments at Sony Sports (Hawk-Eye). Through Kitchen Table AI, I bring that same build-and-run discipline to small businesses: I find the job eating your week, build the tool that takes it off your plate, and stay on to keep it working.

I sell one thing: The Build. Everything below is what it looks like shipped.

Cambridge, MA · Sony Sports · Hawk-Eye · Kitchen Table AI · taking inquiries
Sony Sports · Hawk-Eye
Lead NBA tracking deployments
KinaTrax · acquired by Sony
Led a 15-engineer team
Brown University
Sc.M., Biomedical Engineering
Kitchen Table AI
The Build, live at ScreenTek
The Build · Cambridge / Somerville / Boston

The Build.

A custom AI system, built and run for you. I build the tool that takes a recurring job off your plate, wire it into the software you already use, and stay on to keep it working. Fixed price. You own it.

Fixed fee Built + run by me Wired into what you use You own it

Owners do not have a spare month to figure out which AI tool fits, wire it into the software they already run, and train the team on it. So I do it. I find the highest-ROI automation, build it end to end, and wire it into what you already use: quote requests, order intake, ticket triage, the follow-ups that fall through. You own the result. Not another dashboard to babysit. If a retainer keeps it running, that is me keeping it running.

  1. Discovery, then a fixed price. I map how your business runs and find the automation worth doing first. You leave owning a ranked roadmap and a fixed-price proposal, yours whether you build with me or not. Founding rate for the first three clients.
  2. Build, in two to three weeks. The price is fixed before any code is written. Custom to how you run, not a chatbot bolted on the side.
  3. Embed and run. A week onsite until your team can run it without me, a 30-day grace period, and an optional retainer if you would rather I keep it live on my infrastructure. Month to month, cancel anytime.
  • Your data stays yours. It runs on infrastructure you already use. Where AI calls happen, no training on your data, and fully local if a workflow needs it.
  • You own it from day one. You own the code and the accounts from the start. If we never start a retainer, the tool still works.
Put fifteen minutes on the calendar →
Joseph Stevenson
15
Engineers led at KinaTrax
3
Leagues shipped: MLB · NBA · NCAA
Sc.M.
Biomedical Eng., Brown
Background

Builder. Operator. AI consultant.

By day I lead basketball-tracking deployments at Sony Sports (Hawk-Eye), rolling out the motion-capture systems that record every game in the NBA. Before that I led a 15-person engineering team at KinaTrax, since acquired by Sony, shipping the data pipelines behind that same biomechanical-tracking tech across MLB, the NBA, and NCAA. The thread through all of it: I do not just build systems, I run them in production.

Now I am turning that same in-person, sit-at-your-table approach into Kitchen Table AI, a practice for owners who want AI built into how they actually work. Not another tool to figure out and run yourself. It starts with a conversation at your counter: what actually eats your week, and what I would build to take it off your plate.

The instinct is older than the AI work. I have run a sustainable golf-ball resale business, a seasonal furniture-refurb operation, a rapid-turnaround merch shop outside TD Garden during the 2024 NBA Finals, and an early diffusion-model art venture before the technology hit the mainstream. Different products, same move: spot the gap, ship fast, learn what scales.

Proof

What The Build looks like shipped.

One live in a client's business, one grown into a full product, and a shelf of things I have built for myself. Same builder, same discipline.

The Build, for a client · Live

Live in the wild: ScreenTek.

A 50-year screen-printing shop in Hamden, CT. Live on their site now: a contract screen-printing calculator that lets a buyer price their own job on the spot, any hour, instead of waiting on a callback.

Live · screentekct.com Self-serve quoting Built with AI

Contract quotes used to mean a phone call and a wait, and every wait loses a warm buyer. Now the price conversation starts itself. In build now: the follow-up layer, an email drip that keeps warm leads moving so a quote turns into an order instead of going quiet.

Try the live calculator →

ScreenTek contract screen-printing calculator, live on screentekct.com
ScreenTek calculator pricing a job
A Build, grown into a product

KangRoo: an AI maintenance desk.

A complete working product I designed, built, and shipped end to end: an AI back office for small property managers who run maintenance without an engineering team.

A tenant emails about a leak. KangRoo classifies how urgent it is, picks the right vendor from the manager's own list, drafts the dispatch, chases the scheduling, and logs every decision the AI made to an auditable trail. The manager watches three lanes, Needs You, In Flight, Closed, and KangRoo handles the rest. Email-only on purpose: nothing to install, and it runs alongside whatever software is already in place.

  • Urgency classifier. Reads the email, ranks how serious it is, and shows its reasoning. Correct it once and it learns your call.
  • Vendor dispatch. Picks the right plumber, electrician, or handyman from your list, by job type and availability.
  • Every decision logged. A time-stamped, auditable record from first email to resolution.

Designed, built, and deployed end to end on Next.js, TypeScript, Supabase, Claude, and Resend. Email pipeline verified end to end.

KangRoo manager dashboard: ticket lanes and AI dispatch status
KangRoo ticket detail: AI reasoning trail and vendor dispatch log
Projects · the same builder, off the clock

Built for myself, shipped anyway.

01
Multi-agent command center
ATLAS
The orchestration layer above Claude Code: seven agents, local, file-based. Shipped one day before Anthropic's multi-agent announcement.
Watch →
02
Open-source personal AI OS
ATHENA
A daily-driver that reads across your data and writes the day's playbook. Shipped 16 days before WHOOP announced similar features.
Watch →
03
Co-op RPG for real-life goals
QuestForge
A party-based RPG where finishing real goals levels your character: your workouts are the grind, your life is the quest line. Launched June 2026.
Watch →
04
A closet, cataloged by AI
Side build · July 2026
Photographed everything I own; AI tagged all 166 items, cut them out, and now builds outfits by occasion and color, then names the smartest next buy. Two days, start to finish.
Watch →
05
Career-matching platform
ENGNRS · 2025-26
My first end-to-end AI launch. Shipped it, acquired users, then wound it down: knowing when to kill a product is its own discipline.
GitHub →