Irina Rish

Irina Rish is a Research Professor at Mila - Quebec AI Institute, affiliated with Université de Montréal, and maintains a significant research affiliation with IBM Research. With an extensive publication record spanning decades and continuing through 2025, she is an active leader in artificial intelligence research.

• Primary Affiliation: Mila - Quebec AI Institute, Université de Montréal
• Secondary Affiliation: IBM Research
• Research Focus Areas: Machine Learning, AI Systems, and Applications

Dr. Rish's research interests encompass machine learning theory and applications, with particular expertise in neural networks, language models, reinforcement learning, and efficient AI systems. Her work bridges theoretical understanding with practical implementations across multiple domains including healthcare, time series analysis, and multimodal systems.

Analysis of her recent publications (2023-2025) reveals a strong trend toward efficient AI architectures, with significant contributions to model quantization (particularly ternary models), scaling laws in language models, and novel techniques for continual learning. Her research demonstrates how theoretical insights can translate into practical improvements in model efficiency and performance.

• Key Publication Trends: Model efficiency, scaling laws, continual learning
• Application Areas: Healthcare (EEG analysis), time series forecasting, vision-language systems

Dr. Rish collaborates extensively with researchers across academia and industry, working with notable colleagues including Guillermo A. Cecchi, Djallel Bouneffouf, Eugene Belilovsky, and Matthew Riemer. Her work appears regularly in top-tier conferences (NeurIPS, ICML, ICLR, AAAI) and journals like Transactions of Machine Learning Research, demonstrating both the quality and impact of her research contributions.

She leads research in federated learning techniques, reinforcement learning systems, and multimodal AI architectures, with recent work focusing on making AI systems more robust, efficient, and applicable across diverse real-world scenarios. Her research program demonstrates consistent growth in impact and scope, with increasing publication output in recent years.