ML Infra Engineer: Scalable RL & Post-Training Pipelines

Whitecircle; Paris; Sur place; EUR 90 000 - 140 000; Plein temps

Résumé du poste

White Circle is seeking an ML Infrastructure Engineer to build the systems behind post-training, RL, evaluation, and inference for AI models. You will collaborate with researchers near GPUs, training loops, data control systems, and the infrastructure decisions that influence model learning and product quality.

You will design scalable RL/post-training pipelines, tune end-to-end performance, and implement infrastructure for experiment tracking, evals, dashboards, and agentic development

Qualifications

• Experience designing, building, or maintaining distributed RL/post-training systems at scale.
• Fluency with deep learning frameworks such as PyTorch or JAX.
• Proficiency in Python including concurrency, asynchronous programming, multiprocessing, and optimization.
• Ability to debug distributed GPU workloads across CUDA runtime, containers, NCCL, networking, and storage.

Responsabilités

• Build robust RL and post-training pipelines with quality testing and ablations.
• Design data control systems for data flow through rollout, replay, filtering, evaluation, and policy updates.
• Tune training and inference end-to-end for high throughput across networking, memory, compute, and I/O.
• Investigate how infra choices affect learning dynamics, eval quality, and stability in LLMs and RL.
• Build infrastructure for experiment runs, artifacts, evals, dashboards, and cost visibility.
• Develop agentic development environments including coding-agent harnesses and multi-agent orchestration.

Connaissances

Distributed RL

Post-training systems

PyTorch

JAX

Python

Concurrency

CUDA

GPU clusters

Outils

vLLM

SGLang

TensorRT-LLM

Dynamo

Description du poste

White Circle is seeking an ML Infrastructure Engineer to build the systems behind post-training, RL, evaluation, and inference for AI models. You will collaborate with researchers near GPUs, training loops, data control systems, and the infrastructure decisions that influence model learning and product quality.

You will design scalable RL/post-training pipelines, tune end-to-end performance, and implement infrastructure for experiment tracking, evals, dashboards, and agentic development