NeurIPS 2026 Workshop · Sydney
Seven researchers connecting foundational time-series methods with scientific systems, multimodal data, generative modeling, and reliable deployment.
Confirmed guests
The speaker slate spans academia and industry and brings together model design, benchmark development, scientific machine learning, and production-scale temporal systems.
Professor, Computer Science and Engineering, UC San Diego
Her research develops physics-guided deep learning for spatiotemporal data, including generative models for spatiotemporal sequences and physics-informed neural networks. Her work on multimodal LLM agents and irregular temporal structures engages the workshop's themes of inductive biases and missing-data handling.
Professor of Statistics, UC Berkeley; Director, Big Data Group, ICSI; Group Lead, LBNL; Amazon Scholar
He pioneered randomized numerical linear algebra and collaborated on the Chronos and Chronos-2 time-series foundation models. His work on scaling and transfer behaviour of foundation models for scientific machine learning engages the workshop's themes of scaling laws and evaluation rigor.
Senior Applied Scientist, AWS
He works on large-scale time-series foundation models and leads development of the Chronos family, with prior work on diffusion-based probabilistic forecasting. His work on tokenization-based pretraining and zero-shot transfer engages the workshop's themes of foundation-model design and operationalization.
Assistant Professor, Chemical and Biomolecular Engineering and EECS, UC Berkeley; BAIR; Faculty Scientist, LBNL
Her research develops physics-inspired machine learning integrating physical inductive biases, geometric structure, and differentiable physics, with applications to fluid dynamics and neural PDE solvers. Her work engages the workshop's themes of physics-informed temporal modeling and reliability on continuous-time dynamics.
Associate Professor of Biomedical Informatics, Harvard Medical School; founder, Therapeutics Data Commons
Her group co-developed UniTS, a multi-task time-series model handling classification, forecasting, imputation, and anomaly detection. Her work on multi-task temporal foundation models and biomedical time-series heterogeneity engages the workshop's themes of multimodal temporal learning and reliability in safety-critical domains.
Associate Professor, Department of Data Science and AI, Monash University
He co-authored the MONSTER benchmark for scalable time-series classification and SETAR-Tree, a global tree-based forecasting method. His work on principled model selection and information-theoretic complexity engages the workshop's themes of benchmark realism and evaluation rigor.
Professor of AI and Machine Learning, School of Computer Science and Engineering, UNSW Sydney; Deputy Director, UNSW AI Institute
Her research focuses on data-efficient learning with multimodal sensor data, spatio-temporal modeling, and continual learning, applied to transport and urban systems. Her work engages the workshop's themes of multimodal forecasting and distribution-shift-aware evaluation.
Review leadership
Five invited speakers also serve on the program committee: Rose Yu, Michael W. Mahoney, Abdul Fatir Ansari, Daniel F. Schmidt, and Flora Salim. The full committee is on the Organizers page.