Famealy.ai

Founder · Product, design and build · Exploring seed funding · 2025

I was among the first 0.1% of ChatGPT users and ranked in the top 1% by message volume during its first year.

As agentic coding tools emerged, they filled a gap that had limited me for years. I could finally turn ideas into working software, and my creative output accelerated.

My children arrived in 2023 and 2024. We were a fit couple, but with two young children, food had become an afterthought and kitchen time was limited.

During a small home renovation, eating out kept getting more expensive. We needed a practical way to plan varied meals, cook efficiently and control the weekly shop.

I built Famealy as the solution: one weekly meal plan that adapts to every family member, reduces prep time and combines the household's food into a single optimised shop. It was one of my first substantial software products, and my exploration of agentic tools has not stopped since.

Problem

  • No weekly planWe were deciding what to cook each day instead of working from one family plan
  • Too little varietyAdults had different preferences and children needed a wider range of food
  • Too much kitchen timeCooking needed to stay below one hour a day or be prepared for the week upfront
  • Different nutritional needsEvery adult and child needed meals suited to their own targets and guidance
  • Fragmented shoppingSeparate plans had to become one optimised list rather than several disconnected shops

What I built

  • Custom meal-plan engineA multi-stage algorithm that assigns proteins, searches food combinations, scales portions, fills snacks and enforces allergens, dislikes, price caps and variety rules with layered fallbacks
  • Whole-family planningOne base plan adapts for a spouse and children using individual goals, goal weight, preferences, allergies, calorie targets and macros
  • PrepOrganises ingredients by family member and day, with prep, chopping and seasoning steps, AI-generated flavour variations and food images
  • PlanShows every family member's meals across the week
  • ShopPools the household's ingredients into one optimised shopping list
  • One-click Amazon basketAdds the optimised family shopping list to an Amazon basket in one action
  • AI flavour systemOpenAI turns each base meal into a named, cuisine-inspired variation using its real ingredients, saved into a reusable library spanning 13 cuisines
  • Food-image pipelineGemini generates a consistent, ingredient-accurate image for each flavour variation

Key decisions

  • Start from the account owner's planThe owner manages the primary plan, budget and shopping; it then adapts for adults with close food swaps and for children so the household still cooks the same week
  • Personalise within national guidanceAdults follow guidance for men and women, while children's plans follow age guidance, reuse adult ingredients and generate extras where needed
  • Calculate nutrition before using AIMifflin BMR and TDEE determine calorie needs, followed by diet-specific macro splits and meal distribution with per-meal caps
  • Keep AI inside the nutrition planAdded flavour ingredients can contribute no more than 5% of meal calories; prep steps must reference real ingredients, and generic salt-and-pepper fallbacks are rejected
  • Prevent repeated or competing variationsTwo fingerprints and a similarity score catch renamed or lightly edited duplicates; family cuisine preferences and least use determine ranking, while locks prevent workers racing on the same recipe
  • Generate accurate images onceA fixed reference and style prompt control bowl, angle, lighting and brand colour; exact ingredients, difficult-food hints and cooking methods control accuracy before images are compressed, stored, reused and cost-tracked
  • Test the real generatorForty-three representative families run in dry-run mode inside a locked-down production schema, automatically checking calories, macros, allergens and timing and exposing real defects

Outcome

I built Famealy in approximately four months. A full family plan typically generates in about four seconds, and the app is now in beta through Apple's TestFlight with my family and a small group of friends.

Architecture

Famealy is a monorepo with an Expo app for iOS, Android and web, a shared logic package, Supabase Postgres with row-level security and edge functions, and a separate generator worker running on Railway.