Tristan A.A.
Tristan Ty Aumentado-Armstrong


I am currently a research scientist in artificial intelligence at the Samsung Artificial Intelligence Centre in Toronto.

Gitlab Github Orcid Google Scholar LinkedIn DBLP Twitter


Publications

Display Settings
Sorting
Tags
Descriptions
Images

Bio

Research History

My initial research work was in structural molecular bioinformatics, including macromolecular imaging via transmission electron microscopy and protein surface interface site prediction.

I then spent time in the Computational Systems Neuroscience Lab of Prof. Maurice Chacron, simulating and analysing sensory pathway midbrain cells with invariant recognition properties involved in animal communication. I also interned at the McGill Centre for Integrative Neuroscience (MCIN) within the Montreal Neurological Institute (MNI), working with Prof. Tristan Glatard, on scientific computing tools for neuroinformatics.

Afterwards, I joined the Shape Analysis Group of Prof. Kaleem Siddiqi in the Centre for Intelligent Machines (CIM). I worked on medical computer vision and biophysical modelling of the heart, as well as geometric feature extraction from shapes.

I then entered graduate school, co-supervised by Prof. Sven Dickinson and Prof. Allan Jepson, at the University of Toronto, working on computer vision and artificial intelligence. My thesis encompasses various aspects of shape representation learning, including modelling deformations, weakening supervision, and improving implicit geometry fields.

After my PhD, I was a post-doctoral fellow in the Computational Vision and Imaging Lab (CVIL) at York University, working with Prof. Marcus Brubaker and Prof. Kosta Derpanis on 3D scene inpainting and generative models of multiview image sets.

Currently, I am a senior AI research scientist at the Samsung Artificial Intelligence Centre (SAIC) in Toronto, supervised by Dr. Alex Levinshtein. I have worked on various aspects of computer vision, including neural 3D scene editing, but more recently have focused on generative modelling and super-resolution of images.

Research Interests

My research interests lie in artificial intelligence, particularly computer vision.

My previous work largely focused on representation learning of 3D shape (i.e., how geometry can best be modelled for downstream tasks), with an eye towards disentanglement, weak supervision, inference from images, and generative manipulation. Such approaches can be useful for intelligent agents, such as for intuitive physics simulations or planning in reinforcement learning, but also for creative applications (e.g., 3D scene editing).

In general, I am interested in conditional generative modelling for solving inverse problems in computer vision. My more recent focus has been on image super-resolution, particularly the amelioration of generative hallucinations, while preserving efficiency and image quality.

I am also interested in applying AI techniques to biological modelling and medical informatics (e.g., molecular generation, medical/biological image processing, evolutionary analysis).

Education Background

  • Doctor of Philosophy (PhD) in Computer Science, University of Toronto, 2024
    • Thesis: On Disentangled Analysis-by-Synthesis Shape Representations
  • Masters of Science (MSc) in Computer Science, University of Toronto, 2018
    • Thesis: Geometric Disentanglement for Generative Latent Shape Models
  • Bachelors of Science (BSc), McGill University, 2016
    • Major: Computer Science | Minor: Mathematics
  • Bachelors of Science (BSc), McGill University, 2014
    • Major: Anatomy and Cell Biology | Minor: Computer Science

Contact