|
Maryam AlipourHajiagha I recently graduated with a Master's in Applied Mathematics from Polytechnique Montréal, where I was a graduate research assistant at Mila - Quebec AI Institute under the supervision of Julie Carreau. My research focused on the intersection of probabilistic deep learning, computer vision, and scientific machine learning, with applications in uncertainty-aware modeling and climate modeling. I also worked as a Machine Learning Research Intern at Ouranos, where I developed deep learning models for high-resolution data generation. Before that, I obtained my Bachelor's in Applied Mathematics with a minor in Computer Science from Amirkabir University of Technology, one of the top universities in Iran. My research interests lie in building reliable and interpretable AI systems, especially through probabilistic generative models, uncertainty-aware deep learning, and high-dimensional computer vision pipelines. More recently, I have become increasingly interested in large language models (LLMs), including explainability, reliability, and their integration into end-to-end AI systems. With a strong mathematical foundation and hands-on experience across the full ML lifecycle, I am interested in advancing research on generative AI, multimodal learning, and trustworthy AI systems. |
|
Publications |
|
A Probabilistic U-Net Approach to Downscaling Climate Simulations
Maryam AlipourHajiagha, Pierre-Louis Lemaire, Youssef Diouane, Julie Carreau, NeurIPS 2025 AI4Science Repo / arXiv A Probabilistic U-Net to downscale climate data and compare different training losses. Results show MS-SSIM best captures extremes, while afCRPS best models uncertainty and fine-scale details. |
|
Explainable Generative Models for Downscaling
Maryam AlipourHajiagha, Julie Carreau, B. Gauvin To be announced Details forthcoming. |
Teaching
|