My research revolves around computational optimization, see a non-exhaustive list on the homepage. If solving hard problems, designing and implementing efficient algorithms for convex or mixed-integer optimization, or applying these techniques to concrete applications hypes you, reach out!

Table of Contents

Current Topics

Ph.D.

Doctoral projects are open-ended by nature, the precise topic has to be discussed and adapted to the candidate, so I won’ t have a precise list of current subjects (for now). If you are interested in starting a Ph.D. in Autumn 2024, reach out soon enough (Spring 2024 at the latest) so that we can exchange on your interests and formalize and prepare a topic.

Master’s internships and Master’s theses

The following topics would be suitable for Master’s students, some would be good 1-2 month projects, some could constitute a Master’s thesis / graduation project (6 months).

Method-agnostic reoptimization in convex optimization

See the description here.

First-Order Methods for Constrained Optimization

I am working on the FrankWolfe.jl solver. I have a lot of open threads that can become cool internships or projects:

• a Python wrapper for FrankWolfe.jl
• designing and implementing new Frank-Wolfe variants within the package like the Nearest Extreme Point oracle or a stabilized version of the Blended Conditional Gradients.

Mixed-Integer Convex Optimization

I am working on several methods for optimization methods for mixed-integer problems with a convex relaxation: how to combine techniques from the continuous (convex) and discrete world? You can read about some examples in our preprint describing the methods implemented in the Boscia.jl solver.

Topics for projects will include:

• Applications in sparse statistics, optimal control
• Development of specialized primal heuristics within the Boscia.jl solver.

Mixed-Integer (Nonlinear) Optimization

I am working on SCIP, the largest open-source solver for mixed-integer (non)linear optimization. Projects within SCIP will include cutting planes, heuristics, performance evaluation and presolving. Depending on the type of project, we will probably team up with other people working on the solver.

NB: working on SCIP will mean being relatively confident programming in C and using the surrounding tooling (CMake, gcc).

An efficient parallel network simplex solver

The network simplex is the engine power many applications from combinatorial optimization to Optimal Transport. The goal of the project will be to implement an efficient network simplex solver inspired by LEMON in a modern language (Julia or Rust) that would allow for easier parallelism and portability.

Profiles

I love working with curious and eager learners. Saying that you don’t know XYZ is okay, as long as you are willing to learn. Majors that naturally work well for me are people who studied Applied Mathematics, Operations Research, Computer Science and are already familiar with optimization. Most projects I have will require good programming skills, some will require familiarity with optimization concepts (duality, polyhedral theory, mixed-integer sets, …).

There is an unfortunate bias in the gender of applicants feeling “qualified” for the role. If you feel you belong to an underrepresented group in applied mathematics and computing sciences, I especially encourage you to apply (here and in general), even if when you are unsure about your qualifications.

In general, if you have doubts, reach out and let’s discuss, there might be a project that fits you well and that you hadn’t seen.

Location(s)

I am working in the LIG laboratory in Grenoble and will look for students here.

If you are in Berlin, I can take students in co-supervision with someone else in the IOL lab. See the instructions to apply.
I will be able to supervise a project only if there is a suitable co-supervisor available in Berlin. A project in Berlin can in particular be suitable if you want to work on SCIP.