GeoDataSphere

Inspiration

Mining decisions shape national economies, yet much of the process still relies on fragmented data, paper reports, and expensive manual analysis. In Rwanda, vast geological potential exists, but accessing and understanding that data remains a major barrier for both investors and regulators.

The inspiration for GeoDataSphere came from a simple question: What if geological intelligence were as accessible and interactive as a map?

I wanted to explore how AI and modern web mapping could transform geological data from static documents into a living, explorable system, one that reduces risk, increases transparency, and helps Rwanda attract responsible mining investment.

What it does

Geodatasphere is an AI powered geological intelligence platform that helps users discover, analyze, and evaluate mineral potential across Rwanda.

The platform combines an interactive map with an AI "Director" that understands natural language. Users can ask questions like "Show me high potential gold zones in the west", and the AI automatically navigates the map, activates relevant geological layers, and explains what the data means.

Users can:

Explore satellite, geological, infrastructure, and mineral heatmap layers
Identify high-confidence mineral zones visually
Upload geological pdf reports and extract structured insights ( the administrator account)
Generate professional investment briefs in one click
Assess legal availability and proximity to infrastructure

The result is a faster, cleaner, and more data driven approach to mining exploration. Imagine getting the data as an investor that would have required you to make a deep research or even field visits.

How I built it

GeoDataSphere was built as a web based platform using a modern, AI first approach. And it was entirely build with Gemini 3 in Google AI Studio.

Frontend: React for a responsive and interactive interface.
Mapping engine: leaflet for rendering live satellite imagery, geological layers and heatmaps.
AI layer: Google Gemini 3, Google Gemini 2.0 flash for Natural language understanding, AI driven map control ("Director Mode"), Document reading and data extraction, narrative explainations and summaries.

The architecture was designed so AI does not just answer questions, but actively orchestrate the interface turning the map itself into a Presentation Tool.

Challenges we ran into

One major challenge was coordinating multiple AI driven actions from a single user request. A simple command might require:

Interpreting intent
Selecting map layers
Navigating geographical regions
Generating explanations
Updating visual outputs

Managing this flow while staying within API rate limits required careful prompt design and request sequencing.

Another challenge was presenting complex geological data in a way that remains understandable to non- experts without oversimplifying the science.

Accomplishments that we're proud of

Built an AI-controlled geological mapping platform from scratch
Successfully integrated a natural language control over map visuals
Demonstrated a realistic use case for AI in national resource management.
Designed a system that benefits both investors and government regulators
Showcased how AI can reduce exploration risk and improve transparency

Most importantly, the project shows how emerging AI tools can support real-world economic development, not just experimentation.

What we learned

This project reinforced several key lessons: -AI is most powerful when it controls systems, not just text output

Clear prompt orchestration is essential for multi-step AI workflows
Visual context dramatically improves decision-making
Government, education, and investment platforms benefit greatly from explainable AI

It also highlighted how AI-first design can unlock entirely new interaction patterns.