ML Infra Engineer: Scalable Training & Inference

Mixpeek; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Trajectory is seeking an ML Infrastructure Engineer to join our research-driven team in San Francisco. You will own the innermost loop of our continual learning platform, spanning training, inference and kernels, and collaborate with researchers to translate new algorithms into reliable systems.

We value deep specialization in training infra, inference systems, or GPU kernels, strong Python and C++/CUDA/Triton skills, and the ability to deliver measurable performance gains while preserving

Qualifications

• Strong fundamentals in distributed systems, networking, storage and failure recovery.
• Deep specialization in one of training infrastructure, inference systems, or GPU kernels, supported by measurable optimizations.
• Strong Python skills and languages relevant to your specialty, such as C++, CUDA, or Triton.
• Understanding of PyTorch, JAX, or comparable framework internals, with strong profiling and debugging skills.
• Ownership and clear communication: work closely with researchers and deliver measurable performance gains while preserving correctness and reliability.

Responsibilities

• Build and optimize distributed training and RL infrastructure, including rollout execution, GPU scheduling, training parallelisms, checkpointing, and recovery.
• Improve serving for production agents and training rollouts, balancing latency and throughput.
• Develop and optimize GPU kernels and runtimes using CUDA, Triton, or comparable tools.
• Build reproducible benchmarks, observability, and automated research workflows across training, inference, and kernels.

Skills

Distributed systems

Python

C++/CUDA/Triton

PyTorch/JAX internals

Communication/ownership

Tools

Miles

SkyRL

vLLM

SGLang

ML compilers

Job description

Trajectory is seeking an ML Infrastructure Engineer to join our research-driven team in San Francisco. You will own the innermost loop of our continual learning platform, spanning training, inference and kernels, and collaborate with researchers to translate new algorithms into reliable systems.

We value deep specialization in training infra, inference systems, or GPU kernels, strong Python and C++/CUDA/Triton skills, and the ability to deliver measurable performance gains while preserving