PhD Position F/M Mesh partitioning for high performance solvers on Exascale computing systems

Inria Villeurbanne; Villeurbanne; On-site; EUR 21,000 - 28,000; Full time

Job summary

Inria Villeurbanne invites applications for a PhD position focusing on scalable mesh/graph/hypergraph partitioning for parallel FEM. The project is part of the Horizon Europe FEniCS Lighthouse initiative and involves international collaboration to advance partitioning objectives for scalable simulations.

The PhD will design and implement partitioning algorithms, analyze performance, and work closely with partners to integrate results into the FEniCS framework and related tools, contributing to

Qualifications

• Strong background in HPC and FEM
• Experience with mesh/graph/hypergraph partitioning
• Proficiency in C++ and Python; parallel programming skills

Responsibilities

• Design, analyze and implement mesh partitioning algorithms
• Research optimizing inter‑process communication in FEM
• Collaborate with international project partners and contribute to open‑source FEniCS

Skills

C++

Python

MPI/OpenMP

GPU acceleration

Education

PhD in HPC / computational science

Tools

METIS

ParMETIS

SCOTCH

KaHIP

Zoltan

FEniCS

deal.II

Job description

Context

This PhD will be part of the Horizon Europe HORIZON-JU-EUROHPC-2026-COE-LH-01-03 FEniCS Lighthouse project, which is fully funded and will start in January 2027. The other international partners of the project are Simula Research Laboratory, Norway; University of Cambridge Cambridge, United Kingdom; University College London, United Kingdom; University of Luxembourg, Luxembourg; Koç University, Türkiye; Sabancı University, Türkiye. The PhD will work closely with the researchers in this project.

Assignment

Real-world applications of finite element mesh (FEM) simulations require parallel computing.In such a setting, computational meshes must be partitioned among multiple computers as a first step. One important goal of mesh partitioning is to keep the resulting inter-process communication overhead low while achieving load balance. Failing to do so will lead to parallel FEM simulations which do not scale. The current state-of-the‑art and mainstream approaches treat mesh partitioning as a graph or hypergraph partitioning problem, which is then tackled with well‑known libraries. The common goal of these libraries is to minimize the total volume of inter‑process communication. However, this does not accurately reflect the inter‑process communication cost in the current systems. This mismatch is exacerbated specifically on today’s strongly hierarchical HPC topologies, which have highly variable inter‑process communication cost. The current mesh partitioner needs major improvements to incorporate the proper objectives in its underlying optimization procedures, and thus achieve scalable finite element computations on current and future supercomputers. This PhD will work to develop a novel approach to mesh/graph/hypergraph partitioning by adopting an optimization objective that truly reflects the actual communication overhead.

Main activities

The PhD will design, analyze, and implement algorithms for mesh partitioning and collaborate with the project partners to make use of it in the open source FEniCS framework https://fenicsproject.org.

Skills

Technical Skills:

• Programming: Proficiency in C++, Python; experience with parallel programming models (MPI, OpenMP, or GPU acceleration) is a plus.
• Graph/Hypergraph Partitioning: Knowledge of libraries like METIS, ParMETIS, SCOTCH, KaHIP, or Zoltan, and is a plus.
• FEM Frameworks: Familiarity with FEniCS or deal.II is a plus.
• Experience with supercomputing environments.

Collaboration skills:

• Ability to work in a multi‑institutional, interdisciplinary team (academia + research labs).

Communication skills:

• Clear technical writing and presentation skills (for papers, reports, and open‑source documentation).