Escucha Comunitaria: what we built in six hours in Cartagena

I’d never been to a hackathon with my brother. He’s a political scientist, a social and youth leader, with years of community work behind him. I’m an engineer, working in AI. On paper, the most unlikely pair for a tech event. In practice, the most natural combination we could have brought.

On August 13, 2026, we were in Cartagena for the CTW Hackathon, with one concrete challenge: build something that addresses a real problem in the city. We chose what felt most urgent and most overlooked. The gap between what local governments decide to build, their projects, their policies, and what people actually need.

That gap doesn’t exist because governments are malicious. It exists because genuinely listening to thousands of people is a massive technical problem. And because the information that does get collected, surveys, interviews, community forums, tends to end up in documents that nobody processes well.

Escucha Comunitaria was our answer.

The core idea is simple: what if all that scattered information lived inside a system that could process it, understand it, and respond to it? What if you could ask a corpus of citizen voices what the people of Cartagena think about insecurity, transportation, or healthcare, and get a response grounded in what those people actually said?

That’s what we built. But before I describe the system, I need to talk about how we built it, because that part matters just as much.

While I was in the computer lab, surrounded by participants, writing everything we were going to need, my brother had a different mission: go out into the university, walk through it, and interview as many people as he could find. Ask them about their experiences, their stories, their perception of the city. He was the data source. The one bringing the real voices.

At some point during the day he sent me a message asking why I’d brought him there. He felt like a fish out of water among all the engineers. He couldn’t follow what anyone was talking about, didn’t know what the other teams were doing, and felt completely lost.

I told him not to worry. That he didn’t need to understand anything happening inside. That he’d come with a different mission, and that his mission was harder than mine.

Because getting people to open up is complicated. Even more so in Cartagena, where people are careful about what they say, think about the consequences, and don’t trust easily. It’s not simple to get someone to speak honestly about insecurity, about what doesn’t work in their city, about what the government hasn’t solved. It takes tact, experience, trust. Years of community work. That’s exactly what he brought.

By the end of the day, the system we built was only as good as the data he managed to collect.

A system with three functions.

The first: a conversational brain. Trained on real interviews and surveys from Cartagena residents, on topics they themselves prioritized (insecurity, education, health, transport, among others). You can talk to that brain. You can ask it what the city thinks.

The second: proposal validation. You submit a public policy or a project, and the system compares it against the voices in the corpus. Is it viable? Who mentioned it as a priority? Or, on the contrary, did the same people say they don’t want it, that it isn’t needed? A government that actually wants to listen now has a mirror.

The third was the most complex to implement, and the most interesting. We used unsupervised learning (clustering over the vector database) to find patterns that escape the human eye. Recurring clusters of needs, connections between topics nobody had mapped, structures that emerge when you let AI discover what the data is hiding.

All of this in under six hours. Just two people: a political scientist and an engineer.

Technically, the system integrates generative AI, RAG, advanced data analysis, and unsupervised learning over a vector database. But what matters most to me isn’t the architecture. It’s that it works as a bridge between two worlds that rarely talk to each other: the world of data and the world of people.

We placed second. And before the trip home was over, we were already thinking about how to keep building.

If you want the technical details (architecture, stack, all three functions explained in depth), check out the Escucha Comunitaria project write-up.