news

Jun 15, 2026 Started an internship at Microsoft, working on generative retrieval and exploring how language models can be trained to perform retrieval over large-scale indexes. The work investigates jointly learning representation and retrieval in a single model, effectively compressing the retrieval index into the model weights and reducing storage costs.
May 13, 2026 New preprint out on CHASM: Online Changepoint Detection in Temporal and Cross-Variable Dependence. tl;dr: formalised for vector autoregressive processes, detects changes through the eigenvalues of a recursively estimated transition operator, with a multivariate complex-valued change statistic and successful applications to image and text data streams.
Jul 11, 2025 Our extended paper, Online Changepoint Detection in Multivariate Seasonal Time Series via Dynamic Mode Decomposition, is now under review at a journal, with refined methodology, new theoretical results, and a substantially broader evaluation focused on biomedical signals.
May 24, 2024 New preprint out on Online Changepoint Detection via Dynamic Mode Decomposition. tl;dr: introduces an online, nonparametric approach for multivariate time series, including non-i.i.d. settings with trends, periodicity, or seasonality.