Toward Adaptive Intelligence

Real-world data arrives as an endless stream whose distribution shifts over time. Models that work today may degrade tomorrow, and the human effort required to repeatedly re-tune them does not scale. This talk presents a line of work on adaptive intelligence, where systems select, configure, and revise their own learning pipelines as the data changes. I begin with AutoML for non-stationary data streams, covering online pipeline search, automated drift and outlier handling, and meta-learning for data streams. I then turn to tabular foundation models and ask what happens when in-context learners encounter concept drift. Along the way, I introduce TuiML, an open-source MCP-native ML runtime that enables agents and researchers to run, benchmark, and empirically study machine learning workflows.

Bio

Nilesh Verma is a research fellow at the AI Institute, University of Waikato, New Zealand, working with Albert Bifet and Bernhard Pfahringer. His research focuses on automated machine learning for non-stationary data streams, spanning online AutoML, drift and anomaly handling, and meta-learning for algorithm selection. His PhD thesis on this topic underpins a series of publications at KDD, ICDM, PAKDD, SAC, PRICAI and the AutoML Conference. His current work studies how tabular foundation models behave under concept drift and how to build stream-native foundation models. He is the co-creator of TuiML, an open-source MCP-native ML runtime, and maintains several widely downloaded open-source libraries. He is currently a visiting researcher at Télécom Paris, Institut Polytechnique de Paris.