About

I am a PhD student at McGill University / Mila - Quebec AI Institute supervised by Prof. David Rolnick. I am interested in problems in representation learning and generalization driven by applications in Earth Observation. Some of these problems include:

• Geographic distribution shift: how can we develop models that generalize evenly across geographies of interest, particularly when some are under-represented or unseen during training?
• Learning Earth embeddings: what inductive biases, data sources, and pre-training strategies can yield Earth embeddings that are useful across a wide range of downstream tasks?
• Fusing Earth embeddings: how can we effectively fuse Earth embeddings with other modalities to improve model performance without overfitting?

Prior to starting my PhD, I was employed at Microsoft for 5 years, first as a software engineer and later as an applied scientist. I primarily worked on ML models for satellite imagery and points-of-interest data. Before that, I received my undergraduate degree from the University of Toronto in Math and Computer Science.

I grew up in the Greater Toronto Area and currently live in Montréal. If you are interested in my research or any of my blog posts, please do not hesitate to reach out – I would love to hear from you!

E-mail: rcrasto99 [at] gmail [dot] com