Nutriveat

AI-Powered Personalized Nutrition

A comprehensive health and nutrition platform that leverages fine-tuned generative AI to architect personalized meal plans, automate shopping lists, and provide real-time culinary assistance.

My Role

Lead Developer & Architect

Stack

Flutter, Firebase, OpenAI (GPT-4o), Novita AI, StoreKit / Play Billing

Impact

Fine-tuned LLM Assistants • Direct Store Integrations • Multi-Tier Subscriptions

Interactive Gallery — Select or swipe to explore

System Architecture Log

Traffic Flow

Service Node

graph LR subgraph Client_Mobile [Flutter Frontend] A[Mobile App]:::traffic end subgraph Firebase_Backend [Control Plane] Hub((Firebase SDK)):::hub C[Firestore DB]:::node D[Cloud Functions]:::node end subgraph AI_Orchestration [Intelligence Layer] F[OpenAI / GPT-4o]:::node G[Novita AI / Stable Diffusion]:::node end subgraph Store_Integrations [Native Billing] H[App Store / Play Store]:::traffic end A <--> Hub Hub <--> C Hub --> D D ==>|Fine-tuned Prompts| F D ==>|Image Generation| G F -.->|JSON Parsing| D A <-->|Native IAP Flow| H H -.->|Server-to-Server Hooks| D classDef traffic fill:#2563eb,stroke:#3b82f6,color:#fff classDef node fill:#16a34a,stroke:#22c55e,color:#fff classDef hub fill:#f59e0b,stroke:#d97706,color:#fff

PROJECT LOG // AI ORCHESTRATION // MONETIZATION

The Engineering Story

Nutriveat represents a deep dive into the practical application of Large Language Models (LLMs) in a consumer-facing mobile environment. The goal was to move beyond a standard "chat wrapper" and create a deeply integrated tool that understands the nuance of dietary constraints, kitchen logistics, and user budgets.

Fine-Tuned AI & Structured Output

A major engineering hurdle was ensuring the AI generated valid, consistent, and safe meal plans. I implemented a system of fine-tuned system prompts and strict schema validation within Cloud Functions to force GPT-4o to return structured data. This allowed the app to take raw AI output and instantly transform it into actionable Firestore documents, shopping list items, and high-fidelity image prompts for Novita AI.

Native Subscription Architecture

To support the ongoing API costs of generative AI, I architected a robust multi-tier subscription model (Monthly/Annual). I implemented the monetization layer by integrating directly with the Apple App Store (StoreKit) and Google Play Console (Billing Library). This involved architecting a custom server-side validation system in Cloud Functions to handle real-time subscription status, grace periods, and cross-platform entitlement logic without the use of third-party middleware.

Context-Aware Culinary Assistance

I developed a specialized AI Chatbot designed to function as a "Kitchen Assistant." Unlike general-purpose bots, this assistant is provided with the specific context of the user's current meal plan, dietary allergies, and available utensils. By using RAG-lite (Retrieval-Augmented Generation) principles, the bot can provide accurate unit conversions and tailored cooking instructions that respect the user's specific kitchen setup.