Built with FlutterFlow: Student Innovation Shines at TreeHacks

This February, FlutterFlow sponsored TreeHacks at Stanford University. During this 36-hour hackathon 1,000 college students from across the US worked on innovative projects in sustainability, health, education, Web3, and AI. Of the 260 projects submitted, 25 used FlutterFlow.

Creating for First Responders

Grace Kim, a PhD student in Aero/Astro, created an app that uses satellite imagery to quickly identify areas affected by natural disasters: CHRONOS. Grace used classical machine learning methods to detect changes in real-time satellite images, a Vision Language Model to perform inference and caption the image, and Perplexity’s SonarAPI to provide first responders with more information about the affected region.

Aum Dhruv, Nick Harty, Shamala Chandrappa, and Steven Li built Fhirband, an AI-powered wearable for firefighters that tracks real-time vitals, provides squad-level health insights, and delivers adaptive haptic alerts to enhance safety during emergencies. Their custom-built wearable integrates with Medibound, a unified IoMT (Internet of Medical Things) platform for managing medical devices and receiving AI-driven diagnostics. Medibound is a startup that Aum and Nick have been building — in FlutterFlow — with a mission to put “your health in your hands.”

From Marketplace Template to Award-Winning App

Starting from an e-commerce template from FlutterFlow’s Marketplace, Anna Wu and Elanu Karakus built Lemon, an e-commerce app where consumers and businesses can buy surplus or "imperfect" produce from local farmers at discounted prices to reduce food waste. They built their backend in Supabase, and gave the app a custom color scheme to match their Lemon logo. Their app took home 2nd place in the Hackathon’s Sustainability track for Best Context Incorporation.

Cynthia Wang, Shayaan Sultan, Taarush Grover, and Bernado Melotti won 2nd place in the Hackathon’s Web3 category for Best AI Agents. Their app, MediLedger, allows hospitals to check drug availability at other hospitals without revealing full inventory details, using their decentralized platform. They use Merkle trees, Zero-Knowledge Proofs (zkProofs), and EigenDA to ensure data integrity and prevent stockouts, and used a Blockchain demo app template as a starting point for their UI.

Beginning with a meal plan template app, Roshni Parulekar-Martins, Mahdi Afshari, and Janet Guo built hydRation, an app that uses a phone’s camera to analyze a short video of a user's finger, extracting a PPG (photoplethysmogram) signal to assess hydration levels in real-time and notify users when they need to drink more water. They used Firebase cloud functions to connect with their custom Python backend used to process the user’s video and PPG signals. Their app won 2nd place in the Health Command Center Challenge sponsored by TerraAPI.

Connecting Apps to Wearable Data with FlutterFlow and TerraAPI

Several projects at the hackathon incorporated Terra, who sponsored the Health track of TreeHacks. The Terra API lets apps connect with health data from wearables such as a FitBit or Apple Watch. To access the health data from Terra, most teams connected their FlutterFlow app to a Firebase project, and integrated the Terra API to write data to Firebase.

Team WabiSabi (Viveka Mehrotra, Saanvi Sampada, Ashley Etheridge, Yoon Yati) created a meditation app that uses TerraAPI to read heart rate data from an Apple watch, and play calming audio depending on the heart rate. Their app also included education modules that teach users about meditation using fun and wholesome matcha analogies.

Using an Avengers superhero theme, Wei Han Chua, Caleb Liow, Ashley Saju, and Nor created JARVIS Health, a dashboard designed for military commanders to track heart rate and exertion data from soldiers to identify early signs of over-exhaustion and prevent injuries. To create an animated dashboard that continuously updated metrics such as beats-per-minute, they used Periodic Actions in FlutterFlow’s action flow editor.

Karam Masad created ArmI, an AI-powered fitness coach. ArmI uses reinforcement learning to optimize workout recommendations for each individual, based on their real-time fitness metrics and historical data. Karam developed his backend in Flask, which served as an API layer to handle data processing. He then connected his FlutterFlow project to his custom API endpoints, using Cloudflare Tunnel to securely access his local API without needing to deploy to the cloud.

Connecting FlutterFlow to Custom AI/ML Pipelines

Among the 100+ students who built with FlutterFlow at TreeHacks, a common question was: how do I integrate my FlutterFlow project with a custom AI/ML pipeline? While each team approached the challenge differently, a common strategy emerged: creating a custom API endpoint to handle data processing and/or calls to other APIs and return results—either directly in the API response or by storing them in a database connected to FlutterFlow for further use.

For example, to build VisonMate, an app that uses image recognition and depth estimation algorithms to verbalize obstacles surrounding a visually-impaired person, the creators (Way Zheng, Serena Li, Eva Cullen, Aidan Lee), implemented a custom FastAPI endpoint. This allowed them to quickly make calls to OpenAI, use the Segment Anything and Depth Anything V2 models, and integrate ElevenLabs for text-to-speech, Whisper for speech-to-text, and LangChain for advanced language processing.

Similarily, to create Sprout, an app where kids can upload a drawing, and then engage in a choose-your-own adventure story inspired by their drawing (while also learning vocabulary words for their grade-level), Sirihaasa Nallamothu, Shaan Doshi, and Rhea Rai used a custom FastAPI endpoint to make calls to Imgur API for image uploading, Luma API for image generation, and Gemini Flash 2.0 for story generation, processing inputs & inputs between each call.

To create EcoBite, an app that tracks food waste in a gamified way, Pierre Harbin, Rebecca Combs, Russel Semsem and Ali Fayed created a Flask backend to make calls to Gemini and process data from the United Nations’ Food and Agriculture Official Density Database.

You can find more examples from TreeHacks on DevPost.

It was an inspiring and energetic 36 hours at TreeHacks, seeing students challenge themselves and turn their ideas into impactful, real-world solutions. Here’s to more building in 2025!