SUDS project

Sep 23 2026

Inferring the Invisible: Using AI to Map Dark Matter in the Universe

Research description:

Dark matter makes up most of the matter in the Universe, yet it cannot be seen directly—it interacts only through gravity. This project uses artificial intelligence and data science techniques to uncover the hidden structure of dark matter by analyzing the motion of stars in dwarf galaxies and stellar streams. These small galaxies and elongated star systems act as “gravitational detectors,” responding to the unseen dark matter around them. The SUDS Scholar will apply simulation-based inference (SBI), a cutting-edge approach in machine learning that uses simulated data to train neural networks to infer the physical parameters of complex systems. By comparing simulated and real astronomical datasets, the team will learn how to extract the dark matter distribution and test different theories about its properties.This project offers hands-on experience in modern data-driven astrophysics, combining tools from AI, Bayesian statistics, and computational modeling. Students will gain exposure to real astronomical survey data, explore uncertainty estimation, and contribute to developing machine learning models that help us understand one of the Universe’s greatest mysteries.

Year: 2026

Researcher:
Ting Li, University of Toronto, Faculty of Arts and Science, David A. Dunlap Department of Astronomy and Astrophysics

Students:
Chun On Yu, University of Toronto
Nawaf Bandar Saeed Alrefaie
King Abdullah University of Science & Technology
Yazan Mohammed Bakhshwin,
King Abdullah University of Science & Technology

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Sep 23 2026

Hamiltonian Optimization and Sampling for Deep Learning

Research description:

In the last decade, deep learning models have become an integral part of our lives, from image recognition software to large language models such as ChatGPT. These models are often overparameterized, with many (many!) more parameters than training examples. While this naively implies that these models should just memorize their training data, instead, we find that they generalize extremely well, finding solutions that are often even better than more traditional underparameterized models. This almost-magical ability to generalize rather than memorize turns out to be (in part) the result of how we optimize and sample from these overwhelmingly large models’ parameters. Motivated by this behaviour, this project will investigate new variants of optimization and/or sampling methods based on ideas from Hamiltonian optimization, dynamics, and optics, to see how well they perform across a wide class of problems (potentially including large language models). This will involve a combination of theoretical work as well as empirical studies of how these methods perform on various methods on benchmarks, their stability and dynamics under various conditions, and the implicit and/or explicit regularization that they provide.This project will be co-supervised with Prof. Ricardo Baptista, Department of Statistical Sciences, Faculty of Arts & Science, University of Toronto.

The SUDS Scholar will be actively involved in pursuing a combination of both (1) theoretical work and literature review, as well as (2) conducting empirical studies of how these methods perform (including their stability, dynamics, and regularization properties) across various benchmarks.

Year: 2026

Researcher:
Joshua Speagle, University of Toronto, Faculty of Arts and Science, Department of Statistical Sciences

Students:
Chuxuan Ai, University of Toronto
Yasir Abdullah M Alsugair, King Abdullah University of Science & Technology

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Sep 23 2026

Benchmarking Astronomical Foundation Models

Research description:

Driven by advances in AI, several groups in astronomy are developing large Foundation Models for Astrophysics, large general purpose ML models for performing many tasks. Our group is involved these efforts, in particular connecting these models to natural language models (LLMs). Little work has been done, however, in evaluating the performance of these models. The aim of this problem is to develop a set of benchmarks across a range of astrophysical applications (gravitational lending, galaxy morphology, photometric redshift determination, stellar parameter determination) to test the performance of current and future models.

The SUDS scholar would work on gathering relevant benchmark data sets from the astronomical literature, starting from some that we have already used and then expand to others, write code to run these through existing astronomical foundation models (such as AION-1), and create summary statistics and visualizations of the foundation models’ performance on these benchmarks. Finally, the SUDS Scholar will create an easily accessible resource for others to run the benchmarks on their own models (e.g., sharing it on huggingface).

Year: 2026

Researcher: Jo Bovy, University of Toronto, Faculty of Arts and Science, David A. Dunlap Department of Astronomy and Astrophysics

Students:
Yixuan Cheng, University of Toronto
Reem Omair A Alshahrani, King Abdullah University of Science & Technology

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Sep 23 2026

Multi-view Policy Learning for Surgical Robotic Tasks

Research description:

Previous work from the MEDCVR lab has shown strong results in policy learning for surgical robotic manipulation tasks using methods such as Reinforcement Learning (RL) and Imitation Learning (IL), but most approaches rely on single camera 2D images. However, as the task becomes more complex: the robot must interact in a 3D world requiring spatial reasoning and depth understanding and multi-object interaction, on which single-view learning does not scale effectively. This project will develop methods to capture consistent 3D representations across viewpoints. These representations can then be used to train policies that are more robust to camera viewpoints, and capable of handling inherently 3D tasks such as surgical cutting. This project aims to: Build on previously developed representation learning pipeline; Train policies for 3D manipulation tasks using multiple cameras in simulation; Transfer trained policies to real robots to evaluate performance on real 3D tasks; Compare baseline single-view RL methods to quantify improvements in sample efficiency and robustness to multiple camera viewpoints; and, Publish the findings.

The SUDS Scholar will work closely with the lab’s Amey Pore, Schmidt AI in Science Postdoctoral Fellow. The SUDS Scholar responsibilities will include:

Data collection on surgical robotic simulators based on Unity, Unreal engine and MuJoCo Playground
Data collection on the da Vinci robot and Franka robots
Different camera configuration experiments for Imitation Learning (IL) and Reinforcement Learning (RL)
Developing novel methods for improving existing IL and RL pipeline
Ablation studies and paper writing.

Year: 2026

Researcher:
Lueder Kahrs, University of Toronto, University of Toronto Mississauga, Department of Mathematical and Computational Sciences

Students:
Oluwagbotemi Iseoluwa, University of Toronto
Ammar Salem Alqahtani,
King Abdullah University of Science & Technolog

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more

Sep 23 2026

Classifying Fish Species with Sound

Research description:

The province of Ontario in Canada has one of the greatest densities of lakes in the world. Sustaining and managing these important populations is vital for maintaining the ecosystem and allowing species persistence despite harvesting. A key measure that fisheries managers require is species abundance. This allows them to understand how abundances change spatially and temporally in response to various stressors and to implement effective management strategies. Traditionally, fish population abundances are tracked through invasive capture methods which require time, labour, and material investments and result in the mortality of many fishes.Hydroacoustic surveying has become an alternative to invasive capture methodologies and is currently being tested by the Ontario Ministry of Natural Resources as a possible alternative approach. In this, sonar is used to locate organisms and objects in the water and the sound emitted at up to 400 distinct frequencies bounces off organisms back to a receiver. These signals received may act as a species “fingerprint” allowing the classification of species and abundance calculations. Automating species identification from acoustic responses “remains the ‘Holy Grail’ to acoustic researchers”. Achieving species recognition through hydroacoustic processes will revolutionize the monitoring and management of commercially important fish populations in Ontario and beyond.

The SUDS Scholar will attend weekly meetings with the supervisor; Use a GitHub repository to organize code and data; Write code in python to run deep learning and other machine learning models; and, Prepare presentations on the research.

Year: 2026

Researcher:
Vianey Leos Barajas, University of Toronto, Faculty of Arts and Science, Department of Statistical Sciences

Student:
Guo Yu He, University of Toronto
Osama Tarek Alshabani, King Abdullah University of Science & Technology

Through SUDS, undergraduate students engage in hands-on research focused on data sciences and AI methodology applications.

Learn more