Npv
We're looking for an ML Infrastructure Engineer to join White Circle, an AI Safety company building the policy enforcement and optimization layer for AI systems. Backed by $11M from senior leaders at OpenAI, Anthropic, HuggingFace, Mistral, and DeepMind, White Circle processes 100M+ API calls monthly and runs its own LLMs in production. You will • Build scalable RL and post-training pipelines, including smoke tuning runs for quality testing and ablations. • Design data control systems for rollouts, replay, filtering, evaluation, and policy updates. • Tune training and inference end-to-end for throughput: networking, memory, scheduling, data loading, storage, checkpointing, I/O. • Build infrastructure for model iteration (experiment runs, artifacts, evals, dashboards, reproducibility, cost visibility) and inference infrastructure for post-training and eval loops. • Build agentic development environments: coding-agent harnesses, tool integrations, runtime sandboxes, multi-agent orchestration. Requirements • Hands-on experience designing and running distributed RL/post-training systems at scale (rollouts, replay buffers, reward signals, policy updates, eval loops). • Strong Python (concurrency, async, multiprocessing, performance optimization) and PyTorch or JAX. • Debugging distributed GPU workloads across CUDA, drivers, containers, NCCL, networking, storage, and checkpointing. • Profiling across the stack (py-spy, PyTorch profiler, Nsight, perf, tracing). • Inference stacks: vLLM, SGLang, TensorRT-LLM, Dynamo, or custom serving. • Ability to connect system metrics to model behavior and learning dynamics. • Relocation to Paris (hybrid) required. Bonus • Public builder footprint: open-source contributions to RL, distributed ML, inference, eval, or agent infra; active technical presence on X. • Experience at high-bar AI infra/research teams (xAI, Qwen, ByteDance, Prime Intellect, or similar). • Ownership of custom training frameworks, trainers, schedul