Fernando Torales Acosta
Machine Learning Engineer at Google
I'm an ML Engineer interested in how machine learning can improve the tools we already use. Currently, I work on the AI for Workspace team at Google, integrating advanced models and popular tools like Drive, Docs, and Slides.
Previously, I worked on ML for fundamental nuclear and high energy physics:
- Collaborated with teams to create new ML-based methods for physics experiments.
- Developed differentiable hardware designs optimized with gradient descent.
- Utilized Graph Neural Networks for feature regression and fast transformer-based point cloud diffusion models.
Interests
- ML Engineering
- Model Eval & QA
- Tinkering
- AI for Improving (not replacing) work
Education
- PhD in Physics 2021
UC Berkeley - BS in Physics 2016
SUNY Stony Brook
Experience
Machine Learning Engineer
June 2025 - Present
Google (Alphabet)
- Working on the Workspace AI Platformm team, integrating advanced capabilities into user workflows.
- Improving the underlying intelligence of popular services like Drive, Docs, Sheets, and more.
Postdoctoral Researcher
Feb 2022 - June 2025
Lawrence Berkeley National Lab
- Served as a postdoctoral research fellow in the machine learning for fundamental physics group under Benjamin Nachman.
- Worked directly at the intersection of fundamental physics and deep learning.
- Focused research on detector data deconvolution, differentiable detector design, and generative modeling for collider data.
Doctoral Researcher
May 2016 - Dec 2021
CERN / University of California, Berkeley
- Conducted PhD research as a member of the ALICE collaboration at CERN.
- Identified 'deep photons' by utilizing modern CNNs to capture photons produced in initial hard scatterings.
- Authored thesis studying parton fragmentation via isolated γ-hadron correlations in pp and p–Pb collisions.
