I work on deep learning — writing and training models as well as researching open problems in optimisation and algorithmic fairness. That currently covers spectral clustering, fairness constraints in graph-based learning, and loss landscape geometry.
On the software side I build end-to-end applications and ML pipelines — cross-platform and web applications, fine-tuning SLMs for specific tasks, and training pipelines for computer vision and NLP. I work primarily in Python, Dart, and TypeScript, with Rust for performance-critical components.
