What I Built
EcoSentinel is an autonomous, multimodal environmental monitoring agent. It is designed to act as a "Security Operations Center (SOC) for the Planet," using vision-language models to analyze ecological data—ranging from satellite imagery to local photos of industrial waste—and generate structured remediation plans.
The core objective was to move beyond static data visualization and create an agent that can "see" environmental degradation, reason about regulatory compliance, and remember the history of a specific location to track its recovery over time.
Demo link - https://ai.studio/apps/45fd3395-6b00-4e5a-bb79-ea18db599750
To test it you can try this
Sector: Gran Chaco, Paraguay
Coordinates: -21.432658, -59.578124
Use the image below for the visual analysis
Demo of the same
Code
How I Built It
The project is built on an Agentic Workflow that prioritizes perception, memory, and verifiable identity:
Multimodal Perception (Google Gemini 1.5 Flash): I used Gemini 1.5 Flash via Google AI Studio to process high-resolution images. The model identifies environmental "threats" (e.g., deforestation, oil spills, or illegal dumping) and maps them against ecological standards.
Long-Term Memory (Backboard): To ensure the agent isn't just reacting to a single moment in time, I integrated Backboard. This allows EcoSentinel to store past audit results, enabling "Temporal Awareness." It can compare a new image of a site to one from six months ago to verify if restoration efforts are actually working.
Agent Identity (Auth0 for Agents): To prevent spoofed reports and ensure accountability, I implemented Auth0 for Agents. Every audit report generated by EcoSentinel is cryptographically signed with a unique machine-identity, making the data trustworthy for regulatory bodies.
Structured Reasoning: The agent doesn't just return text; it outputs structured JSON. This allows the data to be instantly piped into a dashboard or used to trigger automated alerts for environmental agencies.
Prize Categories
Best Use of Google Gemini: Leveraged Gemini 1.5 Flash for its massive context window and multimodal capabilities to analyze complex environmental imagery and output structured audit data.
Best Use of Backboard: Integrated Backboard to give the agent persistent memory across sessions, allowing it to track ecological site history.
Best Use of Auth0 for Agents: Used Auth0 to provide a verifiable identity for the AI agent, ensuring that every environmental report is authenticated and auditable.
Created for the Weekend Challenge: Earth Day Edition 2026.
Top comments (0)