Job Title: MLOps Engineer
Company: Physora - A Transvalor Company
Location: Sophia Antipolis
Contract Type: Permanent (CDI)
Status: Executive
Reporting to: Head of R&D
Download the job offer
COMPANY
Physora is a young Transvalor Group company, founded in 2026 and based in brand-new offices at the heart of the Sophia Antipolis technology park. It is dedicated to developing and industrializing AI surrogate models for the numerical simulation of material-forming processes, including forging, casting, welding and heat treatment.
These models can be integrated into the Group’s digital workflow—FORGE®, DIGIMU®, REM3D®, TRANSWELD®, THERCAST® and SIMHEAT®—to reduce cycle times from hours to seconds.
They will also enable us to expand into production and quality control, with real-time prediction solutions powered by high-quality simulations from the Group’s solvers.
POSITION
You will oversee the entire lifecycle of Physora’s models: from building the training infrastructure (scalable, reproducible and traceable) to high-performance industrialization (inference under plant constraints, on-premise and at the edge), including transforming raw data into usable datasets and optimizing architectures through rigorous experimentation, while bringing proven production discipline to avoid past pitfalls and ensure reliable, differentiated and deliverable solutions.
This is the role that determines the pace. It makes the difference between a team running three experiments a week and one running thirty — which, in this field, accounts for most of the gap between success and failure.
MISSIONS
In this role, you will be responsible for:
- Build and maintain the training infrastructure: loading large-scale results, distributed multi-GPU training, experiment tracking, model registry, reproducibility.
- Develop the tooling that transforms raw simulation outputs into usable datasets, including storage organization, schemas and versioning.
- Bring models to the required inference performance level: quantization, export, serving within the plant’s latency budget, on-premise and at the edge.
- Run experiments and ablations, and present their results in a way that informs architectural decisions, rather than simply reporting numbers.
- Contribute to model implementation alongside the modeling engineer.
- Bring production discipline from outside. You will be the person who has already seen these things fail elsewhere, and you are expected to say so.
PROFILE
- Master’s degree or PhD in Computer Science (specialization in Machine Learning / AI) or Computational Physics, combined with significant professional experience in distributed training systems engineering (PyTorch, high-performance computing, large-scale data management).
- Substantial experience with training infrastructure in production, outside an academic context. For this role specifically, this is non-negotiable.
- Deep expertise in PyTorch and its ecosystem: distributed training, profiling, and the debugging experience that comes from having seen models fail silently.
- Comfortable working with large volumes of numerical data and the storage formats that make them usable — HDF5, Zarr, Parquet or equivalent.
- Strong software engineering discipline: testing, version control, code that others can read.
- Curiosity about physics is essential. You do not need to be an expert in forming processes, but you must be eager to understand, decipher and master them.
YOUR STRENGTHS
- HPC and SLURM environments; large-scale distributed training.
- Scientific computing (C++, Fortran, MPI).
- On-premise model serving and embedded inference; MLOps tooling.
POSITION DETAILS
- Location: Biot / Sophia Antipolis;
- Full-time permanent contract (CDI);
- Occasional work from home;
- Annual gross salary: between €52 and €65k;
- Executive status.
OUR BENEFITS
- An international, dynamic, and stimulating work environment;
- A positive atmosphere within teams passionate about digital technologies and scientific research;
- An attractive compensation and benefits package (aide au logement, prise en charge du transport quotidien, RTT, travail à domicile occasionnel).
More about Physora: www.physora.ai