The Story
Most nutrition apps either assign you a meal plan or ask you to search through an endless database. EatyWise works the other way around: the user picks ingredients they actually feel like eating, and AI builds a complete meal around those choices while keeping it aligned with their nutrition goals.
I designed the product architecture, mapped the user flows, built the interface and developed the MVP.
A significant part of the work was backend architecture before any implementation began. The goal was to build a pipeline that would be reliable, maintainable and cost-efficient from the start rather than fix it later. Every calculation that can be handled deterministically happens in code, with reusable functions and a predictable generation pipeline. AI runs only at the step where it genuinely adds something that deterministic logic can't. Fallbacks are in place wherever generation could fail silently.
The result is an architecture already designed for scale — not over-engineered for an MVP, but not built into a corner either.
The current version is being prepared for user validation. The next step is learning from real usage rather than from assumptions.
- •Product design and user flow architecture.
- •Next.js and TypeScript frontend.
- •Reusable backend pipeline with deterministic calculation layer and targeted AI generation.
- •Architecture built for reliability, maintainability and cost-efficiency.
Category
Product