Case study 002
AI-POWERED SAAS
GlucorAI
An AI-powered health product built around real-world data.
Next.js · React · TypeScript · AI APIs · Supabase · Vercel
System type
AI Product / SaaS / Health Data
Role
Product Design · Full-Stack Development
Status
LIVE
The challenge
The challenge
Managing Type 1 diabetes often means jumping between CGM readings, insulin and pump data, meals and carbs, exercise, sleep, health data, and personal notes. The challenge wasn’t collecting more data — it was making existing data understandable in one place with enough context for AI to find relationships a single reading can’t show.
What I built
What I built
A custom user profile and vector database feed AI calls with personal context. An AI carb counter logs food, and generated insights read across the Diabetes Wall — connecting glucose to meals, insulin, activity, and history so the product answers “what can my data teach me?” instead of only “what is my glucose right now?”
GlucorAI is an AI-backed SaaS platform designed to help people with Type 1 diabetes understand the relationships between glucose, insulin, food, activity, and daily events. The core idea was simple: people with diabetes generate an enormous amount of data, but that data is often scattered across multiple apps and devices. GlucorAI brings those signals together — through a custom user profile, a vector database for AI context, an AI food carb counter, and AI-generated insights from the person’s own data.
- Concept
- Architecture
- Implementation
- Deployment
Spec
01
The Diabetes Wall
GlucorAI’s central interface is a unified diabetes timeline. Instead of treating glucose, insulin, meals, exercise, and notes as separate records, the platform combines them into a chronological story — so an elevated reading can be evaluated alongside meals, carbs, IOB, boluses, activity, notes, and historical patterns.
Spec
02
Custom profile + vector context
Each person has a custom user profile so AI calls are grounded in who they are — settings, history, and relevant context — rather than a generic prompt. A vector database stores that context for retrieval, so meals, notes, and patterns can be pulled into later AI calls instead of starting from a blank slate.
Spec
03
AI food carb counter
Users photograph a meal and GlucorAI identifies foods, estimates portions, and returns carb counts — then places that context on the timeline. The carb counter is built to cut logging friction while creating structured signals the rest of the system can actually use.
Spec
04
AI-generated insights
Insights are generated from the user’s own data, not generic advice. The model looks across glucose, meals, insulin, and activity so the product can surface observations like a larger-than-usual lunch rise when insulin on board was relatively low — contextual intelligence, not another chart.
Spec
05
Data Integration & Architecture
Architecture supports CGM, Nightscout, insulin/pump data, meals, Apple Health/activity, notes, and historical glucose — with room for GlucorAI Connect and additional sources. Profiles, vector context, and AI interpretation stay separated from the data layer for scalable iteration on Next.js, Supabase, Vercel, and LLM APIs.
Spec
06
Design, Privacy & Safety
The UI follows Timeline → Context → Pattern → Insight — more consumer product than medical dashboard. Privacy and responsible positioning are core: GlucorAI is an informational tool, not a medical device, and insights do not diagnose, treat, or replace professional advice.
My role
My role
Product Design · Full-Stack Development
End-to-end ownership from concept through architecture, implementation, and production deployment — ai-powered saas · product design · next.js · supabase · health data · ai/llm integration.
Technology
Technology
Next.js · React · TypeScript · AI APIs · Supabase · Vercel
Frontend
Next.js · React · TypeScript
Backend
Supabase · Custom user profiles
Infrastructure
Vercel
Context
Vector database for AI calls
AI
LLM APIs · Carb counter · Generated insights
Architecture
Architecture
- Custom user profile
- Diabetes Wall
- AI carb counter
- Vector context
- AI insights
- Nightscout integration
Key features
Key features
- 01End-to-end AI-powered SaaS product experience
- 02Custom user profiles that carry preferences and history into every AI call
- 03Vector database for retrieval context on AI requests
- 04AI food carb counter from meal photos and logging
- 05AI-generated insights based on the user’s own glucose, meals, insulin, and activity
- 06Unified Diabetes Wall timeline for fragmented health data
- 07Nightscout ecosystem integration and expandable health-data architecture
- 08Mobile-first UI on Next.js, Supabase, and Vercel
Result / Outcome
Result / Outcome
An AI-backed diabetes companion with custom user profiles, a vector database for AI context, an AI food carb counter, and insights generated from the user’s own health data.
Custom
Profiles
AI
Carb counter
User data
Insights
Most diabetes software answers “What is my glucose right now?” GlucorAI is built around a different question: “What can my data teach me?” That shift — from tracking data to understanding data — is the foundation of the product. GlucorAI turns thousands of disconnected health signals into a story people can actually understand.
The build
Screenshots
