DataDiet

DataDiet, a dietary black box and health infrastructure for patients and doctors or users; snap meals, forget them, and surface medically useful patterns and reports when labs or symptoms change.

Comment

Inspiration

When your cholesterol spikes or gut issues emerge, the first question doctors ask is "What have you been eating?" Most patients can't answer accurately, and existing diet apps don't help. They're built for weight loss, not clinical insight. Also, just "eating healthy" may not be true for everyone (for example women have significantly higher gut problems due to food sensitivity and it would be difficult currently to find out what specific food is causing it. There's no currently available data storage for people that contributes to the leading cause of deaths globally. We wanted to build a dietary black box: capture meals effortlessly, track what actually matters medically, and produce doctor ready reports when health signals change.

What it does

DataDiet is both a webapp and an IOS app. The IOS app lets you snap a photo of any meal and forget about it. Our AI identifies the foods and flags health relevant patterns: microplastic exposure from plastic bottles, processed meat consumption (a WHO Group 1 carcinogen), late-night eating (an extreme contributor of bad sleep which directly correlates to life expectancy), high sodium, and ultra-processed foods to name a few. Over time, it builds an insights dashboard showing your dietary patterns. When you need it, one tap generates a professional report correlating your diet with blood work results; ready to share with your doctor.

The webapp offers detailed reports and visuals on your meals, connects you to your primary care doctor (or matches you with one if you have none), and provides a chat interface to talk with an AI about problems you may be having. For example someone having gut problems that day and went to the doctor to get checked can ask the AI "What has been causing my gut problems recently?". The AI will search through your meal data that has already been stored and your blood work data if you uploaded it to provide an expert answer on what foods may have been causing it and why.

How we built it

iOS App: Expo + React Native with Firebase (Auth, Firestore, Storage) and Google Gemini 3 Flash for real-time food analysis
Web App: Next.js 15, Tailwind CSS v4, Framer Motion, with an AI chat feature that lets you query your meal history
Shared Backend: Firebase for cross-platform sync, Gemini for food recognition and report generation

Challenges we ran into

Getting Gemini to reliably detect health flags from food photos required extensive prompt engineering (changing the system prompt). We also had to balance capture speed with analysis accuracy; users expect instant feedback, but thorough analysis takes time. We solved this with a fire-and-forget pattern: save the photo immediately, analyze in the background, update the UI via real-time Firestore listeners.

Accomplishments that we're proud of

Clinically-relevant tracking that goes beyond calories; plastics, carcinogens, meal timing, etc. One-tap doctor reports that correlate diet patterns with blood work. Cross-platform sync between iOS and web auth and data with real-time updates. Onboarding in IOS app lets users customize what signals they want to track (for people with specific problems).

What we learned

Creating a project this ambitious is difficult in the time frame allotted. We also learned that the clinical use case; answering "what have I been eating?", is underserved and genuinely valuable.

What's next for DataDiet

Taking pictures by using Meta Glasses: This is the missing piece in this project that makes it much easier for the user. Simply looking at a meal/drink and asking the wearable to take a picture to then send to our backend, view on phone or webapp is much easier than manually taking a phone out and snapping a picture. Cloud functions for background analysis when the app closes. Integration with health platforms (Apple Health, lab APIs, etc.). Preventing prompt injection in our chat feature. Working on the doctor to patient platform more (implementing things like message feature between the two, and sending the report straight to the doctor on the platform via secure links). Making the app even more visually appealing. Getting users to try the platform and gain feedback on how much it improves their life. Making the reports even better.

Built With

Share this project:

Updates