Two student projects in Saravena show AI applied to local problems, though their validation is still at an early stage

By Comunicaciones UIS · October 16, 2026.

Artificial Intelligence Engineering students at the Universidad Industrial de Santander (UIS), Saravena campus, presented two research projects at the «IA & Ciencia: Investigar con máquinas» event, organized by the Faculty of Sciences. Both projects start from problems in the department of Arauca, Colombia.

The first, by Yazid Esteban Ramírez Arévalo, classifies flood risk on the Cravo Norte river. According to the university, the team gathered some 289,929 raw data points from monitoring stations (flow rates, rainfall and water levels) and cleaned them with the help of AI. It analyzed 3,906 days between 2016 and 2026. Of those, 15% were classified as high-risk, based on a statistical threshold defined in the research itself.

The second, by Juan Sebastián Mateus Mateus, aims to estimate a new student's dropout risk by comparing them with students who previously dropped out or stayed on. This first phase used synthetic data generated with AI, meaning artificial data, to test the model before working with real information. The student cites work, transportation, connectivity, and family and motivational factors as conditions associated with dropping out. His goal is a degree project using real data from the campus.

Our take: the most interesting part is not the technical sophistication, which the source does not detail, but the starting point. These are third-semester students asking questions about their own region instead of repeating generic exercises. That is what makes AI training useful outside the big capitals.

Caution is also warranted. Nothing in the source indicates that the flood model has been validated as an early-warning system. Identifying 15% of days as high-risk in historical data is a first step, not an operational tool. The dropout model, for now, has only been tested with synthetic data, which does not show that it works with real students. In general, closing the gap between an academic prototype and a reliable system requires validation with real data and the involvement of the institutions that must act on it.

Even so, the trend matters. With AI tools within reach of any student, the bottleneck is no longer programming but asking good questions and knowing the terrain. If that capability spreads to the regions, AI can help anticipate disasters and retain talent. It is a small step, but it points in the right direction for the long term.