Happy Tummy

Inspiration

Digestive issues like constipation, gas, bloating, and irregular stools happen all the time for babies (especially 6-24 months during the transition to solids), but tracking what caused it is hard. Parents are busy, tired, and often log things in scattered places and an inconsistent way. When symptoms show up, they're left guessing, and the lack of consistent tracking makes it difficult to spot patterns early and decide what to try next.

We built Happy Tummy to solve this problem. It offers a simple digestion tracker that turns everyday inputs into clear insights and next-step suggestions. We use AI not to diagnose, but to give moms instant and personalized guidance based on the baby's profile and recent digestive patterns. The AI acts like a supportive coach by summarizing what the logs suggest, recommending safe try/avoid food.

What it does

Happy Tummy helps parents:

Create a baby profile (age, birth type, allergies, feeding stage)
Log daily digestion signals (stool type, hydration, gas)
Track food intake using simple food tags and nutrition estimation
Detect patterns between food intake and next-day digestion
Ask questions in a chatbox that gives personalized answers based on the baby's profile.

Instead of diagnosing, our system analyzes trends and associations in daily logs and provides supportive wellness guidance.

How we built it

We built Happy Tummy using:

Expo (React Native) for the mobile frontend
Node.js + Express for a secure AI proxy server
Featherless AI (OpenAI-compatible API) for structured, explainable coach messaging
A rule-based pattern detection engine for constipation trend analysis
A food tagging + nutrition scoring system (fiber, hydration, binding foods)

Challenges we ran into

Running two servers: the Expo app is the client UI, and the Express backend hosts AI endpoints + secrets. This separation keeps API keys off the phone and avoids CORS/network issues.
Food intake tracking: We wanted to go beyond “what food” and estimate how much a baby ate, which meant finding an API or dataset that could support real-world baby foods, portions, and nutrition in a consistent way. Many food APIs are built for adult nutrition tracking, so mapping baby-friendly foods (purees, mixed meals) and portion sizes into meaningful nutrient signals (fiber, hydration-supporting foods, binding foods) took trial-and-error and careful simplification.
UX clarity: parents need fast, calm answers, so we focused on short, action-oriented outputs (3–6 sentences in chat) and instant guidance.

Accomplishments that we're proud of

One accomplishment we’re especially proud of is building a fully working end-to-end AI system in such a short time. Our Expo mobile app connects to a secure Express backend, which integrates with Featherless AI to generate personalized digestion insights and chat responses. We didn’t just plug in a chatbot, we designed structured endpoints (/coach and /chat) that pass baby profile data, recent logs, and controlled recommendation lists into the model to produce safe, contextual answers.

Then, we’re proud that Happy Tummy demonstrates a real preventative health concept. By combining simple daily tracking with AI-powered summaries and instant Q&A support, we built a tool that helps parents notice patterns early and make informed adjustments, before small digestion issues become bigger concerns.

What we learned

Healthcare AI needs boundaries. We learned to write prompts and UX that clearly avoid diagnosis, focus on patterns.
AI is strongest when paired with structure. Instead of asking an LLM to invent everything, we feed it structured inputs (baby profile, insights, try/avoid lists) and constrain outputs to keep results safer and more consistent.