Irina Rish

Irina Rish is an AI researcher whose work spans probabilistic inference, neuroscience and foundation models. After earning a PhD in computer science specializing in AI at UC Irvine in 1999, she spent 20 years at IBM Research, where she worked on projects at the intersection of neuroscience and AI and led the Neuro-AI challenge. She joined UdeM as an associate professor and core faculty member at Mila in 2019, received a Canada Excellence Research Chair in 2020, and secured a DOE INCITE allocation in 2023 for scalable foundation models. In October 2026, she coauthored a preprint that turns TS2Vec into a time-series foundation model. Today, she is a full professor at UdeM, leads the Autonomous AI Lab and is a core academic member of Mila.

Now at: Université de Montréal and Mila
Works: 301
Citations: 8,640
h-index: 34
Badges: Bridge
Loyal

Publication numbers from OpenAlex, as of 27 October 2026.

Timeline

1983

A book on artificial intelligence

In 1983, at 14, she reads a Russian translation of “Can Machines Think?” in Samarkand, Uzbekistan, and becomes interested in artificial intelligence.
1994–1999

Graduate AI training

She earns an MSc in computer science at UC Irvine in 1994 and a PhD in computer science specializing in AI there in 1999.
1999–2019

IBM research and bounded inference

From 1999 to 2019, she works at IBM’s Thomas J. Watson Research Center on neuroscience and AI, and leads the Neuro-AI challenge. In 2003, she and Rina Dechter coauthor “Mini-buckets: A general scheme for bounded inference” in the Journal of the ACM; the paper presents approximation algorithms for probabilistic inference and combinatorial optimization.
2019

Joining UdeM and Mila

In 2019, she joins UdeM’s Department of Computer Science and Operations Research as an associate professor and becomes a core faculty member at Mila.
2020

A Canada Excellence Research Chair

In 2020, she receives a Canada Excellence Research Chair (CERC).
2023

An INCITE allocation on Summit

In 2023, the DOE INCITE program awards her 990,000 node-hours on Summit at Oak Ridge National Laboratory for the Scalable Foundation Models project.
October 2026

Turning TS2Vec into a foundation model

In October 2026, she coauthors “Turning TS2Vec into a Foundation Model for Time Series Classification.” The paper presents TS2Vec-FM, a 1.6-million-parameter convolutional network that reaches 83.3% on UCR-128 and 75.3% on UEA-27 under linear probing.
As of October 2026

Leading the Autonomous AI Lab

As of October 2026, she is a full professor at Université de Montréal, leads the Autonomous AI Lab and is a core academic member of Mila. Her current research covers neural scaling laws and emergent behavior in foundation models, continual learning, generalization and robustness.