← All roles

Member of Technical Staff, Research Engineer, Infrastructure

Fleet
Apply on company site
San Francisco, CA$200k - $600kFull-timePosted 3mo ago

About the role

Research Engineer responsible for owning and optimizing Fleet's GPU training and inference infrastructure, including the training stack, inference stack, research codebase, cluster, and kernel-level components to ensure fast, reliable research at scale.

Responsibilities

  • Own the cluster: Capacity, fault tolerance, failover, and on-call when training stalls.
  • Drive utilization: Profile training runs to identify bottlenecks (NCCL, KV cache, dataloader, allgather) and implement fixes; write kernels when needed.
  • Maintain training and inference: Distributed training (FSDP, TP, PP), rollout generation, async training, and the inference stack the training loop depends on.
  • Maintain the research codebase: Keep it fast and reliable as researchers extend it; refactor without breaking science.

Qualifications

  • Distributed training depth: Hands-on experience with multi-node training (tens to hundreds of GPUs) and major parallelism strategies (DP, TP, PP, FSDP).
  • Performance and kernels: Ability to profile slow training and implement fixes; CUDA and Triton coding; shipped fused ops or attention variants in real training runs.
  • Inference systems: Built or maintained high-throughput inference stacks (vLLM, SGLang); experience with KV cache, continuous batching, speculative decoding, and quantization in production.
  • Research-codebase familiarity: Experience working in sprawling research repos to improve speed/reliability without breaking science; familiarity with RL post-training methods (PPO, GRPO, RLHF) and failure modes.

Skills

Required:Multi-node distributed trainingCUDATritonInference systems
Last updated Oct 15, 2026. Interested? Apply on the company site.

About Fleet

Fleet - AI-powered workflows

Fleet creates simulated worlds and real-world challenges to better understand and shape the behavior of artificial intelligences.

AI-powered workflowsbessemer-ai-infrastructure-2026reinforcement-learning-platforms
Company profileWebsite

More open roles at Fleet