ML Infrastructure Engineer for Scalable RL & Inference

White Circle; Paris; Hybride; EUR 157 701 - 306 642; Plein temps

Avantages offerts par ce poste

Relocation package

Hybrid Paris/London work

Medical insurance

All necessary hardware/tools

Team off-sites

Résumé du poste

White Circle is hiring an ML Infrastructure Engineer to build the systems behind LLM post-training, RL, evaluation, and inference. You will collaborate with researchers near GPUs, training loops, data control systems, and cost-aware infra decisions.

You will work on inference stacks, agentic development environments, and end-to-end pipelines, in a hybrid Paris or London setting with relocation support.

Qualifications

• Experience designing distributed RL/post-training pipelines.
• Proficiency in PyTorch or JAX.
• Proficient Python with async/concurrency.
• Ability to debug distributed GPU workloads (CUDA, containers, NCCL).
• Experience with inference stacks (vLLM, TensorRT-LLM, Dynamo).
• Ability to map system metrics to model behavior and performance.

Responsabilités

• Build scalable RL and post-training pipelines for testing and ablations.
• Design data control systems for training data flow and policy updates.
• Tune training and inference end-to-end for throughput and reliability.
• Investigate how infra decisions affect learning dynamics and eval quality.
• Develop experiment infra: artifacts, evals, dashboards, reproducibility, cost visibility.
• Work on inference infra affecting post-training and evaluation loops.
• Create agentic development environments and tooling.

Connaissances

Distributed RL/post-training systems

PyTorch or JAX

Python (async, concurrency)

GPU/CUDA debugging

Inference stacks experience

System metrics to model behavior

Ownership mindset

Outils

PyTorch

JAX

CUDA

NCCL

vLLM

TensorRT-LLM

Dynamo

Description du poste

White Circle is hiring an ML Infrastructure Engineer to build the systems behind LLM post-training, RL, evaluation, and inference. You will collaborate with researchers near GPUs, training loops, data control systems, and cost-aware infra decisions.

You will work on inference stacks, agentic development environments, and end-to-end pipelines, in a hybrid Paris or London setting with relocation support.