Research Engineer, Code Agents Infra

Sourced from feeny.ai

Mistral is hiring a Research Engineer to build and operate the execution, training, and data infrastructure behind its agentic models and coding assistants. You will design large-scale sandboxing platforms, synthetic data generation pipelines, and distributed RL execution environments across hybrid and multi-cloud clusters. The role sits at the intersection of systems engineering and frontier agent research, working directly alongside AI researchers.

• Design, deploy, and operate a high-throughput sandboxing platform running 1M+ concurrent isolated environments for untrusted LLM-generated code
• Architect and scale pipelines for synthetic code generation, agent trajectories, rollouts, and self-play data collection
• Optimize agent training codebases and distributed runtimes (PyTorch, Ray, SLURM/Kubernetes) to improve GPU utilization
• Reduce sandbox cold-start times using container warm pools, snapshot/restore, and optimized image delivery
• Implement Kubernetes-native controllers, CRDs, and queuing systems for multi-cluster resource allocation
• Ensure strict multi-tenant isolation and security boundaries for untrusted agent code
• Maintain high availability, telemetry, and self-healing across millions of transient jobs, including on-call rotations

• 4+ years in systems engineering, distributed systems, cloud infrastructure, or MLOps supporting LLM/RL workloads
• Experience building high-throughput data processing and generation pipelines (e.g., Ray, Spark, distributed queues)
• Strong Kubernetes and container expertise: custom operators/controllers, Linux cgroups/namespaces, Docker image optimization
• Advanced proficiency in Python, Go, C++, or Rust with a track record of profiling and optimizing ML or backend systems
• Hands-on experience with sandboxing and isolation technologies (Docker, gVisor, Firecracker, WASM)
• Familiarity with task queue systems, resource schedulers, and low-latency queuing architectures

• Experience with lightweight virtualization and microVMs
• Comfort working directly with AI researchers to productionize frontier ideas
• Experience with hybrid and multi-cloud cluster operations