🐍

Mathys E. Jam

PhD Student

Laboratoire d'informatique Parallélisme Réseaux Algorithmes Distribués (LI-PaRAD)

About Me

Currently in my 2nd year of PhD under the direction of Prof Pablo de Oliveira Castro, my research focuses on Input-Aware optimization of HPC kernels. I’m co-advised by Eric Petit, Prof David Defour, and Prof William Jalby.

I’m developping the MLKAPS framework to maximize performance of various industrial applications by leveraging machine learning techniques and genetic algorithms.

Interests

• HPC Kernel Autotuning
• Performance models
• Adaptive sampling

Education

• PhD Input Aware Auto-tuning (Ongoing)

  Université de Versailles Saint-Quentin-en-Yvelines

• MD High Performance Computing and Simulation

  Université de Versailles Saint-Quentin-en-Yvelines

• BSc Computer Science

  Université de Versailles Saint-Quentin-en-Yvelines

📚 My Research

My work focuses on improving performance of HPC applications by using auto-tuning and a variety of Machine Learning techniques.

I mainly contribute to MLKAPS, a tool that automates the generation of decision trees for runtime prediction of design parameters depending on the user input.

Currently, I’m exploring the use of quantile regression to make Bayesian optimization scalable, as well as using bayesian optimization to solve the input-aware autotuning problem.

Featured Publications

Auto-Tuning

MLKAPS: Machine Learning and Adaptive Sampling for HPC Kernel Auto-tuning

This paper presents MLKAPS, a tool that generates decision trees that tune HPC kernels' design parameters to achieve efficient performance for any user input automatically using machine learning and adaptive sampling techniques.

Apr 15, 2025

Recent Publications

Mathys E. Jam, Eric Petit, P. Oliveira, David Defour, Greg Henry, William Jalby (2025). MLKAPS: Machine Learning and Adaptive Sampling for HPC Kernel Auto-tuning.

PDF Code Hal

Recent & Upcoming Talks

No recent or upcoming talks.