GeoNews

Inspiration

Modern news consumption is fragmented and reactive. People receive a constant stream of disconnected headlines, making it difficult to remember past events or recognize long-term patterns. This leads to two systemic issues. The first is News Amnesia, in which audiences forget previous developments and fail to detect emerging trends. The second is Distance Disconnect, where events occurring far away appear irrelevant even when they have real economic or operational consequences.

GeoNews AI was inspired by the need to transform news from passive information into operational intelligence. Instead of simply presenting headlines, the project aims to help users understand how global events propagate through geographic, economic, and political systems.

What it does

GeoNews AI is an AI-powered operational intelligence platform that converts global news into structured, geographically contextualized insights. News events are synchronized in real time, analyzed with language models, and visualized on an interactive world map.

The system generates short explanations of why an event matters, visualizes regional sentiment with heatmaps, and provides a time-based interface that allows users to explore how events evolve. It can also perform deeper analysis to identify potential consequences of global events such as supply disruptions, political instability, or economic shocks.

How we built it

The system integrates several AI services and modern web technologies. News data is retrieved from NewsAPI, processed with language models via the Groq API, and stored in Supabase using a PostgreSQL database. A separate Ollama service performs a deep analysis of selected events.

The frontend is built with React and Vite, using Tailwind CSS for styling. Geographic visualization uses Leaflet for 2D maps and Three.js with react-globe.gl for a 3D globe interface. The backend is implemented with Express, which handles API routing, security middleware, and communication with AI services. Additional natural language processing tasks are handled using Transformers.j, with a DistilBERT question-answering mode, to infer the country referenced in news articles.

Challenges we ran into

One challenge was converting unstructured news articles into structured data that could be mapped geographically. Many articles do not explicitly mention locations, so additional language processing was required to infer geographic context.

Another challenge was producing AI-generated explanations that remain concise while still conveying meaningful implications. Balancing clarity, speed, and cost across multiple AI services also required careful design decisions.

Accomplishments that we're proud of

The project demonstrates how scattered headlines can be transformed into geographically contextualized intelligence. GeoNews AI integrates real-time news ingestion, AI analysis, and interactive visualization into a single platform.

What we learned

The project highlighted the importance of context in AI-driven systems. Information becomes more valuable when it is connected across geography, time, and causal relationships rather than presented as isolated data points.

We also learned that combining multiple specialized AI tools can produce richer insights than relying on a single model.

What's next for GeoNews

Future work will focus on improving the causal analysis engine to better model relationships between geopolitical events, economic indicators, and supply chain networks. Additional data sources, such as shipping information, financial indicators, and satellite observations, could further strengthen the platform.

The long-term goal is to evolve GeoNews AI into a global situational awareness platform that helps organizations anticipate risks and respond proactively to complex global events.