The AfAS Conference Hackathon, held in Kasane, Botswana from 20–22 March 2026, brought together 27 students from diverse African countries and academic backgrounds to tackle a photometric redshift challenge. Hosted by the African Astronomical Society (AfAS) in collaboration with the Inter-University Institute for Data-Intensive Astronomy (IDIA), the Development in Africa with Radio Astronomy (DARA) programme, Hack4dev and the IAU Office of Astronomy for Development (OAD), the event served as a precursor to the AfAS Annual Conference (22–27 March 2026) and marked the fourth annual AfAS Hackathon.
Open to individuals with a strong interest in astronomy or data science, the hackathon welcomed those with basic to intermediate programming skills, particularly in Python and a passion for teamwork and problem-solving. Beyond competition, the event emphasised contribution, encouraging participants to develop solutions that support the growth and advancement of astronomy on the continent. Through mentorship and guided sessions, participants gained practical skills in applying data science techniques within an astronomical context, while also engaging in creativity and innovation through team-based competition for recognition and prizes.
The hackathon challenge focused on developing machine learning models to estimate distances to galaxies and quasars using multi-band photometric data. Participants worked on the ilifu cloud server, provided by the Inter-University Institute for Data-Intensive Astronomy (IDIA), which enabled access to essential computational resources and data. Over the course of the event, participants gained hands-on experience with realistic research problems in modern astrophysics, learning to navigate imperfect and heterogeneous observational conditions.
The hackathon provided students with practical exposure to photometric redshift estimation, galaxies and quasars, photometric data exploration, and machine learning model training. It encouraged original ideas and collaboration, with participants from diverse backgrounds sharing knowledge and expertise. ilifu, the cloud computing platform provided by IDIA, played a crucial role in enabling the hackathon’s success, offering a scalable and reliable infrastructure for data analysis and model training. The hackathon also involved training and validating models on a provided dataset, followed by predictions on a blind test set that included standard objects and more challenging out-of-distribution cases. Through this process, participants worked on developing robust models capable of handling real-world astronomical data conditions.
The teams explored a range of machine learning approaches, with many finding artificial neural networks to be more effective than random forests. The winning team successfully implemented advanced boosting methods, such as XGBoost and LightGBM, to secure first place. A key takeaway was the importance of robustness in dealing with test objects whose properties differed from the training and validation sets.
The winning team, BKN Machines, approached the challenge by treating galaxies and quasars as two distinct problems, rather than relying on a single model to address both. They engineered 258 features from the raw photometric data and trained specialised models for each object type using gradient-boosted decision trees, specifically LightGBM and XGBoost. This targeted approach proved highly effective, achieving a final loss score of 0.019— below the 0.35 threshold required to unlock additional training data from the organisers.
Winning team:
Rubinah Solomon- Botswana Accountancy College
Joy Olayiwola- National Space Research and Development Agency
Magnus Makgasane- Botswana Accountancy College
Seipone Sebina- Botswana Accountancy College
Walter Maketso- Kenyatta University
Beyond the technical outcomes, the hackathon demonstrated the growing capacity of young African scientists to apply advanced data science methods to real-world research challenges, strengthening the continent’s pipeline of skills in astronomy, artificial intelligence, and data-intensive science.
The organising team, Nikhita Ramkilowan (DARA), Nombali Qodi (AfAS), Thembela Matungwa (AfAS/SAAO), Narusha Isaacs-Klein (IDIA/BRICS Astronomy), Theophilus Matsepane (IDIA), Zodwa Tiki (Hack4dev), and Charles Takalana (OAD) prepared and guided the event. Facilitators Ginés Martínez Solaeche (IAA-CSIC), Narusha Isaacs-Klein, Joyful Mdluli (OAD), Nikhita Ramkilowan, and Thobekile Ngwane (SAAO) supported participants throughout the event. Judges Miora Rakototafika, Zolile Mguda, Narusha Isaacs-Klein and Nikhita Ramkilowan commended the creativity shown by participants.
As a precursor to the AfAS Annual Conference, held from 22–27 March 2026, the hackathon set the tone for a week of engagement, knowledge exchange, and collaboration. The hackathon was a resounding success, fostering skills development, collaboration, and confidence among the next generation of African astronomers and data scientists, while reinforcing the role of AfAS and its partners in building a sustainable and globally competitive scientific community across the continent.
Written By,
Zodwa Tiki (Hack4dev)
Narusha Isaacs-Klein (IDIA/BRICS Astronomy)
Theophilus Matsepane (IDIA)