Développeur.se - opérations liées à l’apprentissage automatique / MLOps Developer

Job summary

As an MLOps Engineer, you will optimize machine learning models and build the streaming data path from lake to GPU, ensuring effective operational processes. Key responsibilities include instrumenting GPU utilization and job success rates, as well as managing telemetry for researchers. You will work with technologies such as Python, PyTorch, and AWS for ML workloads.

Qualifications

• 8+ years of engineering experience
• 4+ years building and operating ML systems in production
• expertise in Python
• comfort reading and modifying framework-level code (PyTorch, vLLM, SGLang, TensorRT-LLM)
• demonstrated model or inference optimization work
• hands-on experience with distributed data processing for ML (Ray, Spark, or equivalent)
• practical observability skills (Prometheus, Grafana, or equivalent)
• working knowledge of AWS compute and storage for ML workloads (EC2 GPU instances, S3, EKS)
• experience with containerization and CI/CD for ML artifacts

Responsibilities

• optimize how models train and serve
• build the streaming data path from lake to GPU
• make data formats a deliberate choice
• instrument GPU utilization, queue depth, job success rate, cost per experiment, data-loader stall time, and serving latency percentiles
• land training and experiment telemetry into a trusted location for researchers

Skills

Relocation