SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Mistral is seeking a Research Engineer to join the Forge team, a full-stack AI platform that bridges research and production for enterprise AI systems. You will work end-to-end on model adaptation, post-training (CPT/SFT/RL/distillation), evaluation, data pipelines, and infrastructure—translating customer requirements into reliable, deployable workflows.
Key responsibilities include building and hardening post-training and evaluation workflows, developing tools for synthetic data generation and curation, debugging large-scale distributed ML systems, and improving the Forge codebase through clear APIs, tests, and maintainable abstractions. You'll push the frontier of the RL training stack, ensure seamless deployment across diverse client environments (cloud, on-premises, varied hardware), and partner with researchers and infrastructure engineers to identify and resolve system bottlenecks.
You should have strong Python engineering skills and experience in large codebases (testing, CI, operational ownership), hands-on experience with PyTorch or JAX, and solid systems/infrastructure fundamentals. LLM training or post-training experience (fine-tuning, RL, distillation, evaluation, data pipelines) is essential. You'll need excellent debugging skills in ambiguous, distributed systems and the ability to communicate clearly with both technical and non-technical stakeholders. High agency, low ego, and comfort in fast-moving, under-specified environments are critical.
Nice-to-have skills include distributed training frameworks (FSDP, DeepSpeed, Megatron), cluster orchestration (SLURM, Ray, Kubernetes, Kueue, Karpenter, Skypilot), experience building reliable ML infrastructure or evaluation systems, research background in LLMs/agents/multimodal models, open-source contributions or publications, and experience training multi-billion-parameter models or petabyte-scale datasets. The role sits in Applied Science with direct impact on client outcomes, requiring close collaboration across research, engineering, product, and customer-facing teams.
