Suppertime
An AI kitchen assistant that plans the week, generates recipes from what's already in the pantry, and books a private chef when you'd rather not cook — built for a consumer foodtech startup.
Product preview
Representative interface, rebuilt with fictional data. The client's production UI is not shown.
The challenge
Deciding what to eat is scattered across a dozen surfaces: recipe sites, notes apps, grocery lists, and chef-booking marketplaces that know nothing about each other. The client wanted one conversational product where a household asks a single assistant “what's for dinner?” and gets an answer grounded in its own pantry, preferences, allergies, and schedule — then follows through, whether that means a fifteen-minute recipe, a full weekly menu with a shopping list, or a booked private chef for Saturday. That meant building meal planning, recipe generation, discovery, and a chef marketplace as one coherent system with per-household personalization from day one, and shipping it fast enough for a consumer launch.
Our approach
Made the conversational assistant, Sous, the front door of the product — a guest-usable chat surface that turns “what can I make with chicken, basil, and orzo?” into recipes, weekly plans, and grocery lists, with tool calls into every other product surface
Built the system of record on Supabase — Postgres with row-level security scoping recipes, menus, pantry contents, and bookings to each household, and Supabase Auth with a guest-to-account upgrade path so browsing works before signup
Shipped a Node.js/TypeScript service layer for AI recipe generation and a 10,000-plus recipe discovery catalog with full-text search, dietary and allergen tagging, favorites, and paginated browsing
Delivered a private-chef marketplace — chef profiles, availability, and a session booking flow the assistant can drive end-to-end, with upcoming and archived session management
Results
AI-Generated Recipes in Catalog
Products Unified in One Assistant
Concept to Launch
“We came with a brief that sounded impossible — meal planning, recipes, groceries, and chef booking behind one conversation. alphabench made the assistant the product instead of a feature, and shipped the whole platform in a quarter. People use it every single day, and that was the point.”
Elena Marsh
Co-founder, Suppertime
Capabilities used
Have a similar project? Let's talk.
We'd love to hear about your consumer AI or platform engineering challenges. Let's talk architecture.
START A PROJECT