EcoEquity

Orchestrating satellite telemetry, real-time vision, and voice-driven AI to resolve urban thermal inequity

Comment

🌍 Inspiration

Urban heat islands are not just an environmental issue — they are a matter of social justice.

Many low-income communities suffer from:

🌡️ Higher temperatures 🌿 Lack of vegetation 🏥 Increased health risks

Yet, the data to understand these problems (satellite imagery, NDVI, thermal maps) is:

Complex Technical Not accessible to everyday people

👉 EcoEquity AI was born to democratize geospatial intelligence Transforming raw satellite data into real-time, voice-driven insights anyone can use.

💡 What it does

EcoEquity AI is a multimodal geospatial AI platform that:

🛰️ Detects Urban Heat Islands (UHI) using satellite data 🌿 Analyzes vegetation health (NDVI) in real time 🎙️ Allows users to talk to the map using voice 🤖 Uses AI to explain, predict, and recommend actions 🚀 Key abilities: “Go to the hottest zone in Casablanca” “Why is this area at risk?” “Suggest solutions for this neighborhood”

👉 The system responds live, moves the map, highlights zones, and gives actionable strategies.

🧠 How we built it

🏗️ Core Stack: Frontend: Next.js + TypeScript + Tailwind CSS AI Core: Google Gemini (1.5 Flash, 1.5 Pro, 2.0 Flash, 3.1 Preview) Cloud: Google Cloud Vertex AI Satellite Data: Sentinel-2 (via Sentinel Hub) Voice: Web Speech API + SpeechSynthesis State & Storage: Zustand + Supabase ⚙️ Architecture Flow: 🛰️ Fetch satellite imagery (Sentinel-2) 🧠 Process NDVI + heat anomalies 👁️ Capture map frames via Canvas API 🤖 Send to Gemini (multimodal reasoning) 🛠️ Execute AI function calls: Move map Highlight risk zones 🎙️ Return voice + visual feedback ✨ Challenges we ran into ⚡ Real-time performance with large satellite data 🎙️ Synchronizing voice input + AI + UI actions 🧠 Making Gemini understand visual map context 🌐 Ensuring accurate environmental data (grounding) 🔁 Managing streaming responses (SSE) smoothly

👉 Solved using:

Vertex AI streaming Function calling Smart state management Multimodal pipeline optimization 🏅 Accomplishments that we're proud of ✅ Built a fully interactive AI agent inside a map ✅ Real-time voice → action → visualization loop ✅ Integrated satellite + AI + UI control ✅ Created a system usable by non-experts ✅ Designed a futuristic geospatial UX

👉 This is not a demo — it's a working intelligent system.

📚 What we learned

Multimodal AI is powerful when connected to real-world data Voice interfaces make complex systems accessible Satellite data can drive local social impact AI agents are more useful when they can act, not just answer

🚀 What's next for EcoEquity AI

🌍 Deploy globally (city-level analysis worldwide) 📱 Mobile version for field workers 👥 Community reporting system (citizen feedback loop) 🧑‍⚖️ Government dashboards for urban planning 🌳 AI-powered automatic reforestation planning

Built With

Share this project:

Updates