Inspiration
My friend Anthony Williams cuts hair as Antz Cuts, a one-man barbershop in Pennsylvania. Watching him work, I saw him lose money two ways: no-shows that leave his chair empty, and hours of back-and-forth DMs just to set up a time. I'm a CS student, so I built him the app I wished he had. His customers can go from "what cut should I get?" to booked and paid in under a minute, and he doesn't have to touch his phone.
What it does
- Face-scan Style Finder. Snap a selfie and the app maps your face shape, then recommends the cuts that fit it. One tap on "Book this cut" goes straight to booking with the service picked.
- AI booking assistant. Type it like you'd text a friend: "fade Friday after 5" or "kids cut for my son tomorrow." Google Gemini works out the service, day and time. My code then matches that to a real open chair, so the AI can never double-book or invent a slot. It also answers questions about prices, hours and the membership.
- PayPal pre-deposits. Every booking is locked in with a $5 PayPal deposit that comes off the cut. The booking only confirms after PayPal reports the payment complete for the exact amount. No-shows now cost the customer, not Anthony.
- Unlimited Haircuts Membership through PayPal. Regulars pay $85 once through PayPal for 30 days of unlimited haircuts and skip deposits. On the 5th cut it's still free when they bring a friend, so members become referrals.
- Installs like an app. It's a Progressive Web App with the Antz Cuts logo, a home-screen icon and offline-ready pages.
Try it: https://antz-cuts.onrender.com ยท Code: https://github.com/ronwiafejr23/Ant-s-cuts-webpage-
How we built it
I built it solo:
- PayPal (sandbox): PayPal JS SDK buttons in the browser, with Orders v2 create + capture on a Node.js server. The secret key and all prices live server-side, so nobody can edit the page and pay $0.01.
- AI booking: Google Gemini (Flash-Lite, with Flash as a backup when it's busy) using structured JSON-schema output. Gemini gets the shop's live services, prices, hours and membership policy. It returns booking intent, and my JavaScript turns that into a real time slot.
- AI Style Finder: Google MediaPipe Face Landmarker runs fully in the browser. It finds 478 face landmarks, measures face length and forehead, cheek and jaw widths, classifies the face shape, and maps it to cuts.
- App + hosting: HTML, CSS and vanilla JavaScript, a zero-dependency Node.js server, a service worker and web manifest, deployed on Render with a one-click
render.yamlBlueprint.
Challenges we ran into
- Keeping the AI honest. Early on, letting the model pick times produced confident but impossible slots. Splitting the job fixed it: Gemini extracts what the customer wants, and code picks a real open slot.
- Gemini's schema rules. Testing against the live API showed Gemini rejects empty options in a response schema, which would have broken every chat message. I caught it before launch and switched to an explicit "none" value.
- Money safety. I never trust a price from the browser. Every PayPal capture is checked against the expected amount before a booking or membership counts.
- Secrets. I kept the PayPal secret and Gemini key out of the browser and out of GitHub (
.envlocally, environment variables on Render).
Accomplishments that we're proud of
- I built this for a real friend's real business, and it solves problems he has every week.
- Face scan โ AI booking โ PayPal payment in under a minute, live on the web.
- An AI that can't double-book, because it never decides availability on its own.
- A membership model that pays up front through PayPal and turns Anthony's regulars into referrals.
What we learned
How PayPal's create โ capture order flow works end to end, how to get reliable structured output from an LLM, how to read face geometry from MediaPipe landmarks, and how to turn a plain website into an installable app. I also learned that building for a real person beats building for a hypothetical user. Anthony's actual hours, prices and pain points shaped every feature.
What's next for Ants Cutz barber app
- A bookings database so Anthony sees every appointment on his phone
- PayPal subscriptions so memberships renew automatically
- Text reminders before each appointment to cut no-shows even more
- Letting the AI assistant suggest a cut from the Style Finder result