Speaker
Description
Radio astronomy is reaching a point where discovery is limited less by telescope sensitivity than by the intelligence of the systems interpreting the data. As MeerKAT and the Square Kilometre Array produce data at unprecedented scale and complexity, machine learning is no longer an optional technique but a scientific instrument in its own right. This keynote argues that code and trained models now sit alongside antennas and correlators as core components of discovery. Embedded within calibration, radio-frequency interference mitigation, source detection, and survey validation, machine-learning systems shape scientific outcomes and encode assumptions that must be understood, tested, and trusted. By integrating physical insight, uncertainty, and interpretability into automated pipelines, radio astronomy can ensure that intelligent models extend—not replace—scientific reasoning, enabling scalable, transparent, and transformative discovery.