SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mistral is building full-stack AI solutions spanning frontier models, developer tools, applications, and compute infrastructure. The company partners with enterprises across finance, manufacturing, defense, healthcare, and the public sector to co-create customized AI systems.
As a Research Software Engineer on the Sandboxes team, you will design, build, and operate the platform that runs isolated environments for untrusted code at massive scale. This is a software engineering role focused on distributed systems, low-level virtualization, and product-minded platform design—not ML research. You do not need prior ML experience; what matters is writing clean, reliable systems code that scales. You will join a small, senior team.
Key responsibilities:
- Design, build, and operate the sandboxing platform running untrusted code at scale across multiple Kubernetes clusters, from bare metal data centers to hyperscalers
- Develop the sandbox runtime leveraging Firecracker, enabling fast startup, snapshotting, and efficient scheduling
- Obsess over performance and latency optimization to reach millions of concurrent sandboxes
- Write robust, well-tested systems code in Rust and Python; review code, CI/CD, and observability across the platform
- Own cross-cluster networking, security boundaries, isolation guarantees, and monitoring for suspicious activity
- Collaborate with researchers and ML engineers who use sandboxes for synthetic data generation and RL training, translating their needs into platform features
Requirements:
- Strong background in software engineering and distributed systems, with experience building and operating large-scale infrastructure
- Deep expertise in at least one of: Kubernetes and cluster operations, cross-cluster networking, observability, or low-level virtualization
- Fluency in Python or Rust; comfortable writing systems code where performance and security matter
- Strong software-design instincts: modular code, testing, CI/CD, observability, and a bias for reliability
- Self-starter who learns fast; core skills matter more than specific tools
- Low-ego, collaborative, and comfortable with ambiguity in a small, high-ownership team
Nice-to-haves:
- Experience with Firecracker, microVMs, or similar virtualization technologies
- Product-mindedness: understanding of how sandboxes are used for synthetic data generation or RL training
- Experience with GPU clusters, CUDA, or ML workloads
- Prior SRE or platform engineering experience at scale
- Open-source contributions or widely used internal tools
