Copa Pulse

The 2026 World Cup's 48 teams hit 16 cities. CopaPulse uses Amazon Nova Lite to analyze news feeds, creating sentiment heat maps. Find the best spots to cheer based on real-time geolocation data!

Comment

Inspiration

With the 2026 World Cup expanding to 48 teams across an entire continent, the scale is unprecedented. We realized that fans wouldn't just be in stadiums; they’d be flooding city squares, local landmarks, and fan zones across three countries. We wanted to solve the "Where is the party?" problem by creating a real-time pulse of the tournament that helps fans find their tribe and the best atmosphere, regardless of which of the 16 host cities they are in.

What it does

CopaPulse is a real-time sentiment and "fan-density" engine. It scrapes global and local news via RSS feeds, filtering specifically for World Cup and football-related updates. Using AI, it identifies:

Geolocation: Which specific landmark or district is being discussed.

Sentiment: Whether the vibe is celebratory, tense, or electric. It then visualizes this data on a dynamic heat map, allowing fans to see exactly where the strongest "spirits" are located in real-time.

How we built it

The core of the project is powered by Amazon Nova Lite on AWS Bedrock.

Ingestion: We built a pipeline to ingest massive streams of data through RSS functionality.

Multimodal Analysis: We leveraged Nova Lite’s multimodal capabilities to process text and related media at a fraction of the usual cost.

Filtering & Extraction: We prompted the model to strictly ignore non-football news and extract structured JSON containing location, sentiment, and nationality data.

Mapping: We integrated a tool-calling layer that takes these coordinates and updates a front-end heat map via a mapping API.

Challenges we ran into

Filtering noise was our biggest hurdle. During a World Cup, "football" appears in everything from politics to advertisements. And also connecting it to multiple sourcing channels is still something to be updating. Fine-tuning the prompts to ensure Amazon Nova Lite only processed "boots-on-the-ground" fan sentiment, and required significant iteration. We also had to manage the latency of processing hundreds of simultaneous RSS updates to keep the heat map "live."

Accomplishments that we're proud of

Having a concept come to live that we think could have real use real fast by many people in the World Cup but also any big event that brings many cultures together, and for people as us that love to travel and find different ways to connect, this is another avenue to connect to live events and their people like never before.

What we learned

We have learned that the multi-modal agentic system is not enough to making a working or useful software, it has to have very specific to the results it wants to accomplish, or in other words, go deeper instead of wider. In the case of Copa Pulse, going deeper made us learn more about the important API connections we still need to accomplish bewteen the sources and the processor ( Nova), as well as the filtering process to make it actually serviceful for people using it.

What's next for Copa Pulse

There is still much refinement to be done, and something we talked in depth was about not only sourcing and identifying the heatpoints per city, also sourcing per specific neighbourhoods or zones in those cities. For example in Mexico City, when a big match is going on bewteen 2 teams, their fanbases tend to separate and reunite in different zones, so we want a heatmap that can pinpoint you to those parts of the city depending on the team you are rooting for.

We are very grateful with the opportunity to try Amazon Nova, and to continue our journey as developers.

Built With

Share this project:

Updates