Oceanguard

Inspiration

Ocean pollution is not just an environmental issue — it’s a data coordination failure. Monitoring systems are reactive, fragmented, and slow. By the time communities detect a crisis, the damage is already spreading.

We built OceanGuard to explore a new paradigm: What if AI systems could predict environmental damage before it arrives — and autonomously coordinate human response?

What it does

OceanGuard is a multi-agent AI environmental intelligence platform that:

Aggregates real-time coastal data

Predicts debris and kelp movement 6–48 hours in advance

Ranks cleanup urgency using AI analysis

Verifies site conditions using a conversational phone AI

Transparently funds operations via on-chain blockchain transactions

It transforms ocean monitoring from reactive reporting into predictive coordination.

How we built it

OceanGuard orchestrates five AI systems into a unified pipeline:

OpenAI → Structured reasoning + environmental summarization

Google Gemini 2.0 Flash → Streaming environmental reports

Snowflake Cortex AI → Predictive analytics + phone agent reasoning

ElevenLabs → Human-quality voice synthesis

ExecuTorch → On-device drone inference

Twilio → Real-time speech recognition + call routing

Solana → Transparent on-chain donation tracking

The architecture uses:

React + Three.js for a real-time 3D environmental globe

Node/Express backend with SSE for live transcript streaming

PostgreSQL + MongoDB + Snowflake for structured and analytical storage

Background data engines refreshing every 30–60 seconds

The key innovation is orchestration — coordinating multiple AI APIs with different latencies and behaviors into a seamless human-facing system.

Challenges we ran into

Synchronizing multiple AI APIs with different response times

Streaming live phone transcripts while running predictive background engines

Maintaining real-time updates without bottlenecking the frontend

Designing a human-in-the-loop system that balances automation with oversight

Accomplishments that we're proud of

Built a full AI pipeline: detection → prediction → verification → dispatch → funding → reporting

Implemented a multi-turn conversational phone agent that verifies environmental conditions live

Created predictive debris movement visualizations using animated 3D arcs

Connected environmental intelligence to blockchain-based financial transparency

This is not a dashboard. It’s a coordinated AI system.

What we learned

Multi-agent orchestration is harder than building individual AI features

Real-time AI systems require careful latency management

Environmental AI requires balancing prediction with human trust

Blockchain transparency increases accountability in impact-driven systems

What's next for Oceanguard

Scale beyond 20 cities

Move Solana integration to mainnet

Expand drone edge AI inference capabilities

Introduce autonomous route optimization for cleanup fleets

Deploy in partnership with environmental NGOs

Built With

• executorc
• framer
• google-gemini-2.0-flash
• openai-gpt-4o-mini
• react
• react-globe.gl
• snowflake-cortex-ai
• solana
• tailwind-css
• three.js
• twilio-(voice-+-speech-recognition)
• typescript