Francisco Emiliano
López Saavedra
MSc student in Computer Science, Université de Montréal & Mila -reinforcement learning and machine learning for climate.
I work on machine learning for physical and scientific systems. The methods that model these systems well tend to be expensive, so they stay with the institutions that can afford the compute. I'm interested in making them available to everyone, so that anyone who needs the answer can get it.
Currently I build AI surrogates for wildfire simulators, cutting runtime and computation cost while modelling next year's wildfire hazard from real-world data.
reinforcement learning · ML for climate · surrogate modelling · responsible AI · open science
Super-resolving coarse climate model output to local grids, comparing convolutional and residual architectures.
Deep reinforcement learning policy driving a quadruped from onboard vision alone, using an intrinsic reward to actively explore unfamiliar terrain.
Grounding vision-language model instructions into low-level manipulation actions through a Vision Transformer-based policy.