EcoEngage

Inspiration

Environmental awareness platforms often struggle with declining engagement and high moderation overhead. While users care about endangered species, there is little structured incentive to consistently create meaningful content. Moderators must manually review posts, detect spam, and assign rewards.

EcoEngage was inspired by the idea that AI agents should not just generate responses, but reliably take action — automating rewards, moderation, and accountability within environmental communities.

What it does

EcoEngage is an AI-powered blogging platform focused on endangered species awareness. Users can create posts, comment, and engage with content while earning tokens for meaningful contributions.

An AI Token Agent monitors platform activity in real time, evaluates content relevance using search and scoring, awards tokens automatically, detects potential spam, and executes reliable actions such as updating balances or triggering moderation workflows. Every action is logged and explained to ensure transparency.

How we built it

The platform uses a full-stack architecture:

Frontend: React / Next.js with Tailwind for the blogging interface, authentication flow, and token dashboard.

Backend: Node.js with Express handling APIs, JWT authentication, and Elasticsearch integration.

Elasticsearch Serverless: Stores users, posts, and token balances; enables vector search for relevance scoring; uses ES|QL for querying and updating data.

Elasticsearch Agent Builder: Powers the AI Token Agent using built-in search tools, ES|QL queries, and workflows to execute multi-step reliable actions such as rewards, moderation tickets, and redemption logic.

The agent follows a structured process: retrieve context, reason over relevance and activity, execute updates, and log explanations.

Challenges we ran into

Designing reliable automation without encouraging spam was a key challenge. The agent needed multi-step validation before awarding tokens.

Integrating search, structured queries, and workflow execution required careful orchestration to ensure state consistency.

Maintaining explainability while enabling autonomous decision-making also required deliberate logging and validation design.

Accomplishments that we're proud of

Successfully implementing an agent that performs real, verifiable actions rather than simple text responses

Automating token rewards and moderation logic through structured workflows

Building an explainable reward system backed by search-based relevance scoring

Designing a scalable architecture that reduces manual moderation effort

What we learned

We learned how to design agentic systems that combine reasoning with execution.

We gained practical experience using Elasticsearch for both semantic search and transactional updates.

We also learned the importance of explainability and validation when building autonomous systems that impact user incentives.

What's next for EcoEngage

Next, we plan to:

Improve fraud detection using more advanced similarity thresholds and anomaly detection

Introduce partnerships with environmental NGOs for real-world reward redemption

Expand the agent to support multilingual conservation content

Add analytics dashboards to measure environmental engagement impact

EcoEngage demonstrates how AI agents can reliably take action to increase engagement, ensure accountability, and create measurable environmental impact.