ML Platform Engineer — Training & Inference

Physical Superintelligence; Boston (MA); Híbrido; USD 140 000 - 190 000; Tempo integral

Resumo da oferta

Physical Superintelligence is building AI-powered infrastructure to enable researchers to train and deploy models at scale. You will turn research code into production systems, implement robust training and inference pipelines, and own services used by researchers daily.

You will specialize in distributed training, inference serving, or the ML platform while collaborating across lanes to ship reliable, scalable solutions for cutting-edge physics discovery.

Qualificações

• 3+ years building and operating ML training or inference infrastructure in production.
• 3+ years of software engineering on systems others depend on, with contracts and tests.
• Depth in at least one lane: distributed training, inference serving, ML platform, or gateway infra.

Responsabilidades

• Turn research code into scalable systems with contracts, tests and on-call ownership.
• Build services researchers use to launch runs and route across models.
• Own the training and inference infrastructure: distributed jobs, GPU scheduling, model serving on the shared platform.
• Go deep in one lane and contribute to others on the platform.

Conhecimentos

Software engineering

ML infrastructure

Distributed training

Inference serving

Ferramentas

PyTorch DDP

FSDP

Ray

vLLM

SGLang

Triton

Kubernetes

GPU scheduling

Descrição da oferta de emprego

Physical Superintelligence is building AI-powered infrastructure to enable researchers to train and deploy models at scale. You will turn research code into production systems, implement robust training and inference pipelines, and own services used by researchers daily.

You will specialize in distributed training, inference serving, or the ML platform while collaborating across lanes to ship reliable, scalable solutions for cutting-edge physics discovery.