Skip to main content
  1. Blog/

ActInSpace 2022 — AI Disease Detection from Satellite Imagery

Rauf Ibishov
Author
Rauf Ibishov
Based in Heilbronn, Germany. Building retrieval and ranking systems at Azerbaijan’s National AI Center. Starting my MSc at TUM Heilbronn — looking for werkstudent roles in NLP, search, or ML infrastructure.

Key lesson: with four people and 24 hours, the bottleneck is deciding what NOT to build. The first two hours of scoping discipline are what got us to a working demo.

In November 2022, I led a team of four at ActInSpace 2022 — an international hackathon organized by the French National Centre for Space Studies (CNES) and the European Space Agency (ESA), held simultaneously across 34 countries and 66 cities with 2,900+ registrants and 410+ teams competing worldwide. Our local edition took place at UFAZ (French-Azerbaijani University) in Baku — local partner: Azercosmos — with 32 teams competing over a 24-hour sprint.

What we built
#

Our challenge sat at the intersection of health informatics and space technology: use satellite imagery to detect and predict disease outbreaks. The idea was that environmental signals visible from space — vegetation changes, water body patterns, land surface temperature anomalies — correlate with conditions that enable disease spread (mosquito breeding grounds, contaminated water sources, crop blight).

We built an AI-based detection model that processed satellite imagery to identify environmental risk markers. The 24-hour constraint forced us to keep the architecture simple and focus on a working demo over a polished pipeline.

What I learned
#

This was my first time leading a team at a hackathon rather than just contributing code. The main lesson was about scope management under extreme time pressure — with four people and 24 hours, the bottleneck isn’t coding, it’s deciding what not to build. We spent the first two hours narrowing from a broad “disease detection” idea to a specific, demonstrable pipeline, and that discipline is what got us to a working demo by the deadline.

It also introduced me to working with geospatial data and remote sensing — a domain I hadn’t touched before. The experience of applying ML to a completely unfamiliar data modality in 24 hours was a useful stress test of how transferable ML engineering skills actually are.


  • Event: ActInSpace 2022 — CNES/ESA International Hackathon
  • Venue: UFAZ (French-Azerbaijani University), Baku
  • Local partner: Azercosmos
  • Role: Team lead (4 members)
  • Format: 24-hour hackathon, 32 teams locally, 410+ teams across 34 countries / 66 cities globally

November 2022 — Baku, Azerbaijan


Related#