GeoDude

Inspiration

Mosquito-borne diseases like dengue, malaria, and Zika affect hundreds of millions globally, yet predicting and understanding mosquito habitat patterns remains challenging. We were inspired by the urgent need for accessible geospatial intelligence tools that public health professionals and researchers can use to identify breeding grounds, predict disease hotspots, and allocate prevention resources more effectively. With 45,000+ field observations available, we wanted to unlock the hidden patterns in this data.

What it does

GeoDude transforms raw mosquito habitat data into actionable intelligence:

Interactive Geospatial Explorer - Visualize mosquito observations on a global heat-map. Allows visualization of the correlations between Google Earth-Engine specific categories (i.e., elevation, temperature, precipitation, and vegetation) and mosquito/land-cover data.
AI-Powered Analysis Engine - Ask natural language questions and get expert insights on species distribution, geographic patterns, and temporal trends.
Dataset Dashboard - View statistics, browse records, and identify key patterns at a glance.
Multi-Query Support - Run predefined analyses (General Overview, Geographic Patterns, Species Distribution, Temporal Trends) or ask custom questions.

How we built it

Frontend: Streamlit for rapid web UI development.
Data Processing: GeoPandas for geospatial operations, Pandas for data manipulation.
Visualization: PyDeck for interactive mapping, Plotly for charts.
AI Integration: OpenAI GPT model for intelligent analysis.

Challenges we ran into

JSON Serialization of Geospatial Data - Timestamp and geometry objects aren't natively JSON-serializable. We solved this by converting datetime columns to ISO format strings, and dropping the geometry column before sending it to OpenAI.
Streamlit + Geometry Columns - Streamlit's dataframe display utility doesn't handle GeoSeries objects, producing warnings. We resolved this by dropping geometry before display while preserving it for mapping.
Large Context Windows - Sending 43K records to an LLM would exceed token limits. We engineered smart summarization: top species, country distributions, water source statistics, and sample records to give comprehensive context without bloat.

Accomplishments that we're proud of

Functional AI Integration - Successfully integrated GPT analysis with geospatial data, bridging two traditionally separate domains.
Demo Mode - Created fallback functionality that displays statistics without requiring API keys.

What we learned

Domain + AI = Insight - Combining geospatial data with AI analysis creates genuinely useful outputs that neither tool alone could produce.
Context Engineering Matters - How you structure data for an LLM dramatically affects response quality; summaries > raw dumps.
Environmental Data Complexity - Mosquito ecology involves multiple interconnected variables (elevation, water type, species, temporal patterns) requiring nuanced analysis.

What's next for GeoDude

Advanced Filtering - Filter by elevation range, date range, species, water source type for targeted analysis.
Export Capabilities - Save analysis reports as PDF/JSON, download filtered datasets.
Predictive Modeling - Train models to predict ideal conditions for specific mosquito species.
Integration with Real-Time Data - Connect to live field survey data feeds.
Disease Risk Mapping - Layer epidemic data to correlate habitats with known disease transmission zones.