한국어·English
Mistral AI company · OPEN ROLE
- Company
- A French AI company that builds frontier AI models, developer tools, and compute.
- Role
- Refines AI model training and deployment processes based on customer needs.
Log in to mock apply, save this role or see more roles.
What you will do
We rewrote the responsibilities and qualifications in plain language instead of copying the original posting.
Responsibilities
- Turns real customer requirements into reliable training and deployment workflows.
- Builds and improves post training and evaluation workflows (CPT, SFT, RL, distillation) into repeatable recipes.
- Builds the tooling and pipelines behind synthetic data generation, data curation, and training, plus evaluation and deployment support.
- Debugs and hardens large scale ML systems, including distributed training, scheduling, and checkpointing.
- Improves the Forge codebase with clear APIs, tests, and documentation to keep it maintainable.
- Partners with researchers and infrastructure engineers to turn bottlenecks into concrete system improvements.
What you need
- Has strong Python engineering skills from working in large codebases, including testing, code review, and CI.
- Has hands on experience with PyTorch, JAX, or similar.
- Has experience with LLM training or post training, such as fine tuning, RL, distillation, evaluation, or data pipelines.
- Can debug ambiguous systems, including distributed jobs, data issues, quality regressions, and infra failures.
Good to have
- Has distributed training experience with FSDP, DeepSpeed, Megatron, or similar.
- Has cluster or orchestration experience with SLURM, Ray, Kubernetes, or similar.
- Has built reliable ML infrastructure, evaluation systems, or large scale data pipelines.
- Has research experience in LLMs, agents, multimodal models, reasoning, or code.
- Has experience training multi billion parameter models or petabyte scale datasets.
Read the employer’s original job description
See how your resume reads for this role
A former Korea GM at a global company reads it and tells you what to improve for this role. It uses 1 credit.
Checked 2026-10-18