Technical Manager – Scientific Computing, HPC & GPU Software · MOX Laboratory
Paolo Joseph Baioni
Leading the design and operation of high-performance computing infrastructure and scientific computing technical activities at the MOX Laboratory (Politecnico di Milano), supporting research across applied mathematics, physics, and engineering.
Focusing on GPU-accelerated scientific software, performance portability, and modern C++ for heterogeneous architectures (CUDA, ROCm), including the design and optimisation of numerical methods and simulation codes across CPU/GPU systems, bridging research needs with efficient and maintainable implementations.
Background in computational physics and numerical methods, with experience in massively parallel simulations, particle methods, and plasma modelling workflows. Contributions include peer-reviewed publications and scientific software development, as well as serving as a reviewer for Computational Particle Mechanics (Springer).
Key Skills
Focus on GPU optimisation, parallel algorithms, and development of portable, high-performance scientific software for research applications.
Research & Application Domains
Experience in scientific software development and optimisation, as well as physics-based modelling and large-scale simulations, combining numerical methods with high-performance computing techniques.
Experience
2023–Present
Leading HPC infrastructure development and scientific computing strategy. Designing GPU-enabled systems and developing high-performance scientific software, while coordinating technical activities and supporting research projects.
2022–Present
Delivering MSc-level laboratories in scientific computing (C++, MPI, GPU programming) and supervising students on HPC and GPU-related topics. Co-supervision of a PhD Student.
2021–2025
Developing massively parallel numerical methods and GPU-accelerated simulation codes with performance portability across architectures.
Selected Projects
Design and deployment of modern GPU-based HPC systems to support research workloads.
C++20 + TBB and (roc)Thrust implementation of a massively parallel Material Point Method targeting CPU/GPU architectures with performance portability.
Plasma edge modelling for nuclear fusion using SOLPS-ITER. Background strengthened through participation in the Max Planck Institute for Plasma Physics Summer School.
Implementation of a neural network from scratch in C++ using Eigen and STL.
Repository →Links
Contacts
paolojoseph dot baioni at polimi dot it
paolojoseph dot baioni at gmail dot com