Alain Tissier

CSE Student @ ETH Zürich · Embodied AI · CUDA / HPC

Bio

Motivated ETH Zurich student in Computational Science (specialising in Robotics). Combines hands-on experience in Embodied AI, HPC and CUDA with strong analytical and technical skills.

Topics of Interest

Embodied AI, High-Performance Computing, CUDA / GPU Computing, Optimization, Probabilistic Machine Learning, Simulation, Numerical Methods.

some of my work

Simulated Unitree G1 humanoid tracking a reference motion

Teleoperation for Humanoids

Training motion-tracking policies that let a Unitree G1 humanoid follow human reference motions in simulation, as a foundation for real-time teleoperation (ETH Robotics Club).

In progress

Cyclone Franklin tracks predicted by AI weather models compared with ERA5 and IBTrACS

Causal Inference in AI Weather Models (BSc thesis)

Investigating whether AI weather models such as GenCast, GraphCast and Atlas-CRPS capture physically causal relationships, using tropical cyclones as case studies against ICON, ERA5 and IBTrACS.

In progress

GPU-accelerated linear programming solver visualization

GPU-accelerated LP solver (group project)

Developed a CUDA-based revised simplex solver and evaluated GPU acceleration trade-offs on sparse linear programs.

ETH Robotics Club reinforcement learning project

Reinforcement Learning for Robotic Control

Implemented and trained reinforcement learning policies for robotic control tasks, focusing on stable learning, reward design, and sim-to-real robustness.

Useful resources

Learning material I've put together — not part of my active project work.

ETH Zürich main building (Hauptgebäude)

Studying Computational Science (CSE) at ETH Zürich — a student's guide

A practical, honest guide for prospective students: what the Computational Science / Rechnergestützte Wissenschaften bachelor is actually like, the workload, the specialisations, and advice from inside the program.

DDPM · CFM
DDIM
ViT · DiT

What is the difference between Flow Matching and Diffusion?

A visual reference deck for modern diffusion / flow-matching image generation: training frameworks (DDPM, Flow Matching), sampling methods (DDPM sampler, DDIM) and architectures (ViT, DiT), with the core equations and how they combine in models like Stable Diffusion, SD3 and Flux.