← back to portfolio
case study· 2025· Applied AI Engineer· SEED Hackathon 2025

GaiaNet

Biodiversity intelligence, powered by multimodal AI.

🌿
Overview

GaiaNet turns fragmented wildlife data — images, audio, surveys — into ranked, actionable conservation decisions inside a single dashboard.

Challenge

Conservation teams have data but not decisions. Species IDs sit in one tool, population trends in another, and action-planning happens in spreadsheets.

Outcome

Built and demoed end-to-end during SEED Hackathon 2025, going from raw wildlife data to a prioritized action list in a single flow.

Approach

How it was built.

  1. 01

    Used Gemini 2.5 Pro as the reasoning core, with CLIP zero-shot classification as a fallback for unseen species.

  2. 02

    Forecasted population trends with tabular models and modeled ecosystem stability from combined signals.

  3. 03

    Ranked conservation actions by impact-per-effort so field teams see what to do first, not just what's wrong.

  4. 04

    Shipped as a dark-mode Streamlit dashboard with pydeck maps and one-click PDF field reports.

Stack
StreamlitGemini 2.5 ProCLIPHuggingFacepandaspydecklibrosafpdf2
GitHub →← All projects