Senior AI Training Infrastructure Engineer

Kodiak; Mountain View (CA); On-site; USD 190,000 - 260,000; Full time

Benefits offered by this job

Equity

Annual bonus

Medical/Dental/Vision

Flexible PTO

Office perks: lunch/kitchen

Headspace/Calm services

401(k)

Commuter benefits

Job summary

Kodiak Robotics, Inc. in Mountain View, CA is seeking ML systems engineers to design and optimize high‑throughput data loading and distributed training infrastructure for autonomous vehicle AI workloads. You will enable multi‑node GPU clusters, mixed precision, and scalable dataset pipelines.

The role requires strong Python and PyTorch skills, plus systems-level experience in C++/CUDA/Triton. Competitive compensation and comprehensive benefits are offered.

Qualifications

• Experience with distributed training frameworks and techniques (PyTorch DDP/FSDP, DeepSpeed, Megatron, NCCL).
• Experience building high-performance data pipelines for large-scale training (streaming formats, sharding, storage/network-aware loading).
• Deep understanding of GPU performance: mixed precision, memory hierarchy, kernel fusion, profiling tools (Nsight, PyTorch Profiler).
• Strong Python skills and PyTorch internals; systems-level experience (C++/CUDA/Triton) a plus.
• Passion for scalable AI infrastructure for fast, reliable training.

Responsibilities

• Design high-throughput data loading and streaming systems for multimodal sensor data, including dataset formats, sharding, and prefetching pipelines that keep GPUs saturated.
• Build and optimize distributed training infrastructure across multi-node GPU clusters, applying various parallelism strategies.
• Maximize utilization of accelerators (BF16/FP8), memory optimization, fused kernels, and overlap of compute and communication.
• Profile end-to-end training pipelines to identify bottlenecks across storage, network, CPU preprocessing, and GPU compute.
• Develop scalable dataset construction pipelines converting raw driving logs into training-ready formats.
• Partner with ML teams to scale architectures from prototype to full-cluster training runs.

Skills

Distributed training

PyTorch DDP/FSDP

DeepSpeed

Megatron

NCCL

Data pipelines

Python

C++/CUDA/Triton

Education

BS/MS/PhD in CS or related field

Tools

WebDataset

MosaicML Streaming/MDS

NVLink

InfiniBand

Job description

Kodiak Robotics, Inc. in Mountain View, CA is seeking ML systems engineers to design and optimize high‑throughput data loading and distributed training infrastructure for autonomous vehicle AI workloads. You will enable multi‑node GPU clusters, mixed precision, and scalable dataset pipelines.

The role requires strong Python and PyTorch skills, plus systems-level experience in C++/CUDA/Triton. Competitive compensation and comprehensive benefits are offered.