Senior/Staff Software Engineer, Infrastructure (ML)

Nimble Robotics•San Francisco, CA

About The Position

We’re looking for a Software Engineer to join our ML Infrastructure team. In this role, you’ll help build the training and inference systems that power our general-purpose warehouse robots. You’ll own training infrastructure end to end: keeping GPUs highly utilized, making runs reproducible, and ensuring every researcher can launch the next experiment with a single command. You’ll work closely with ML and Robotics teams to design, build, and scale the systems that turn our GPU clusters into a reliable, high-throughput platform for model development.

Requirements

Bachelor’s, Master’s, or PhD in Computer Science or a related field, or equivalent practical experience.
4+ years of industry experience in infrastructure, distributed systems, ML systems, robotics, or a related area.
Experience with programming languages such as Rust, Go, Python, or C++.
Experience with ML frameworks such as PyTorch or JAX.
Strong understanding of distributed systems, systems programming fundamentals, memory management, and performance optimization.
Experience with Kubernetes orchestration, resource scheduling for large distributed jobs, and containerized deployment pipelines.
Ability to debug and optimize bottlenecks across GPU memory hierarchy, networking fabric, filesystems, and multi-GPU operations.
Ability to reason from first principles and optimize systems for both memory-bound and compute-bound workloads.
Strong cross-functional communication skills, ownership, and a growth mindset.

Nice To Haves

Hands-on experience with distributed training frameworks and techniques such as PyTorch DDP/FSDP, DeepSpeed, Megatron, or NCCL.
Hands-on experience with GPU kernel development.
Experience with data engineering technologies such as Parquet, Arrow, or similar systems.

Responsibilities

Design, develop, and maintain ML training infrastructure that enables the AI team to run training jobs efficiently, manage and iterate experiments quickly.
Build low-latency inference pipelines for production robotics workloads.
Develop, tune, and optimize low-level CUDA kernels.
Design training-platform systems for scalable model training, including high-throughput data ingestion, dataset sharding and sampling for distributed training.
Participate in and lead design reviews with peers and stakeholders to evaluate technical tradeoffs and select appropriate technologies.
Review code and provide feedback to uphold best practices around style, correctness, testability, performance, and maintainability.
Contribute to documentation and educational materials, adapting content as systems and workflows evolve.
Mentor junior engineers and help raise the technical bar across the team.