Ramón Calvo González
I’m a PhD student in Computer Science at the University of Geneva, advised by François Fleuret and funded by Meta.
I’m currently a research intern at Cartesia, working on LLM post-training and adaptive computation for image encoders with David W. Romero and Albert Gu.
Previously, I completed my MSc in Robotics, Systems and Control at ETH Zürich as an ESOP scholar. My thesis at the Robotics Systems Lab explored safe reinforcement learning for wheeled-legged robots. I also interned at NVIDIA, working on sensor calibration and synthetic data, and at Sony R&D, developing compact vision transformers for event cameras.
Outside research, I enjoy learning to play the guitar, cooking, and weightlifting.
Publications
Laminating Representation Autoencoders for Efficient Diffusion
Ramón Calvo-González, François Fleuret
World Model Workshop, MILA, 2026
FlatDINO compresses DINOv2 representations into 32 latent tokens, reducing the computation needed for diffusion-based image generation.
Leveraging the True Depth of LLMs
Ramón Calvo González*, Daniele Paliotta*, Matteo Pagliardini*, Martin Jaggi, François Fleuret
Preprint and Foundation Models in the Wild Workshop, ICLR, 2025
* Equal contribution
Parallel evaluation of consecutive transformer layers improves LLM inference throughput without retraining, with a small accuracy trade-off.
Trajectory optimization and control of a free-floating two-arm humanoid robot
José L. Ramón, Ramón Calvo González, Adrian Trujillo, Jorge Pomares, Leonard Felicetti
Path generation and control of humanoid robots during extravehicular activities
José Luis Ramón, R. Calvo, A Trujillo, Jorge Pomares, Leonard Felicetti
Selected projects
DINO-JAX
End-to-end DINO implementation in JAX, including evaluation and multi-device distributed training.
URDF Editor
Cross-platform C++17 robot editor with interactive 3D link and joint manipulation, full undo/redo, and no ROS dependency.
Tiny WBC
A weighted whole-body controller implemented in C++.
Scientific Calculator
A scientific calculator built from scratch in my final year of high school, powered by an ARM Cortex-M0+.
Updates
ICML Gold Reviewer
Recognized for contributions to the ICML 2026 peer-review process.
Joined Cartesia as a research intern
Working on LLM post-training and adaptive computation for image encoders with David W. Romero and Albert Gu.
New preprint: Laminating Representation Autoencoders for Efficient Diffusion
Introducing FlatDINO, a compact latent representation for efficient image generation.
Leveraging the True Depth of LLMs published in TMLR
Our work on parallelizing transformer layers for faster inference.
Earlier updates
Started a PhD in Machine Learning
Under the supervision of Prof. François Fleuret, funded by Meta.
Started a Software Engineering Internship
At NVIDIA's Autonomous Vehicles team, working on computer vision and Omniverse.
Started working at the Robotics Systems Lab of ETH Zurich for my master’s thesis.
Researching end-to-end safe locomotion and navigation policies for wheeled-legged robots.
Won 1st at the Ideathon for a Novel Sustainable Packaging Material
The ideathon was organized by the Student Biolab of ETH Zurich and Nestle.
Started as a Research Engineer Intern at Sony R&D Zurich.
Worked with event camera algorithms for autonomous driving.
Started a MSc in Robotics, Systems and Control at ETH Zurich
My studies are funded by the Excellence Scholarship and Opportunity Program of the ETH Foundation.
Completed my BSc in Robotics at the University of Alicante
Graduated with an Excellence Distinction (first of class)