ChefMii

A marketplace connecting clients with verified private chefs. A Gemini-powered assistant handles discovery, quotes and bookings end to end - private dining made as easy as ordering takeaway.

Comment

Inspiration

I got the inspiration from my old sister who was looking for a Ghanaian private chef in London but did not find one as there were not any of them, so I decided to build one.

Hiring a private chef is something most people assume they cannot do. The market leader proved the demand exists but the model still relies on human sales teams, human support, and days of back-and-forth. I kept asking: what if AI did all of that? Then a private chef costs no more effort than ordering takeaway, and every skilled chef gets a full calendar without spending a penny on marketing. ChefMii is that bet: a private chef marketplace that a single founder can operate, because AI runs the operations.

What it does

ChefMii connects clients with professional private chefs for home dinners, events, and corporate occasions. Clients browse vetted chef profiles, chat with the ChefMii Assistant, a Gemini-powered concierge that answers questions about chefs, cuisines, pricing, and packages, and guides them from first enquiry to confirmed booking and then book and pay securely through the platform. Chefs get a dashboard, bookings, and client messaging. The platform takes a 10% commission on bookings.

How we built it

Next.js 14 full-stack app deployed on Vercel. The ChefMii Assistant calls the Gemini API (gemini-2.5-flash via the official Google SDK) live in production from our /api/chat endpoint and it is the front line of the business, doing the work of a support and sales team. Firebase Authentication and Cloud Firestore handle users and data, Stripe handles payments, Resend handles transactional email. I built solo, using AI agents (Google Antigravity and Gemini) as my engineering team. AI did not just go into the product, it built the product.

Challenges we ran into

The biggest was migrating the entire data and auth layer to Firebase close to the deadline, 25 files touched, with a hard rule: a working app beats a perfect stack. Scoping was the other fight. The codebase had ambitions (marketplace, academy, multiple dashboard types) and I had to cut ruthlessly to ship a focused, working booking flow. And the classic marketplace cold-start problem: solving supply and demand at once, alone, on a clock.

Accomplishments that we're proud of

Shipping a production marketplace as a solo founder in weeks and I have Gemini assistant, real auth, real payments infrastructure, real waitlist. 0 users acquired with £0 marketing spend. And proving the core thesis to myself: one person plus AI really can operate what used to take a team.

What we learned

AI-native is not a feature, it is a cost structure and every workflow AI handles is headcount the business never needs. I learned to treat scope as the enemy: judges, users, and founders all prefer one working flow over five half-built ones. And I learned that the hardest part of a marketplace is not code at all, it's earning the first real users, which no AI can shortcut.

What's next for ChefMii

Convert the waitlist into first paid bookings in London and onboard an initial cohort of vetted chefs. Deepen AI operations: AI-generated chef menus and profiles to make onboarding instant, and an AI booking concierge that turns "tapas for 12, Mum's 70th, two vegetarians" into a shortlist with real quotes. Then incorporate, raise pre-seed, and scale city by city and an AI-operated marketplace with structurally lower costs than any incumbent.

Built With

Share this project:

Updates