SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mistral AI is building full-stack AI solutions from frontier models to developer tools and compute infrastructure. The Data Infrastructure team is architecting the backbone for frontier model training and fine-tuning at massive scale.
In this role, you will be a core contributor to designing and operating next-generation data infrastructure. You'll take full lifecycle ownership of projects spanning architecture, implementation, and production operations. Key responsibilities include:
- Building and scaling massive distributed compute and storage systems to support world-class AI model development
- Architecting multi-cluster orchestration layers to optimize workload placement across diverse hardware and regions globally
- Designing modern storage formats and systems to handle fine-tuning datasets at exabyte-scale
- Contributing to the internal training platform, ensuring seamless model training capabilities across Kubernetes and SLURM environments
- Implementing metadata and lineage systems to provide visibility as data and model pipelines grow in complexity
- Managing cloud-native deployments and ensuring operational excellence at scale
- Participating in on-call rotations for critical training jobs
You should have 4+ years of experience in Data Infrastructure, MLOps, or Infrastructure Engineering. Strong proficiency in Python is required. You'll need deep expertise with Kubernetes-native tooling and experience debugging large-scale distributed systems. Experience with modern columnar storage standards and foundational compute/storage platforms is highly valued. The ideal candidate is comfortable with ambiguity in a rapid-growth AI environment and takes pride in building reliable, secure, scalable systems from the ground up.
The role is based at one of Mistral's European offices (Paris, London, Warsaw, or Zurich) with a hybrid arrangement. Remote EU/UK candidates are considered with a requirement for one hub visit per month and one week onboarding visit (covered).
