AI Performance Engineer: Scale ML Throughput & Efficiency

Decisive Point; Sunnyvale (CA); Hybrid; USD 180,000 - 240,000; Full time

Job summary

Applied Intuition, Inc. is seeking a performance engineer to optimize large-scale machine learning workloads in datacenters. You will focus on distributed training across many nodes and high-throughput inference over vast autonomy logs.

The role centers on throughput, goodput, and cost per data unit, with ownership of profiling and bottleneck resolution. The ideal candidate has hands-on ML performance engineering experience, multi-node training scale familiarity, and strong Python/C++ skills.

Qualifications

• Hands-on ML performance engineering: profiling, roofline, throughput optimization, root-cause investigation.

Responsibilities

• Profile and optimize distributed training end to end: data loading, preprocessing, kernel execution, gradient communication, checkpointing.
• Optimize offline/batch inference on petabyte-scale logs: batching, quantization, execution, and saturation across long-running sweeps.
• Establish roofline models and quantify gaps between achieved and theoretical performance; rank opportunities by impact and effort.
• Improve multi-node scaling efficiency: sharding, parallelism, collective communication, memory bandwidth, kernel fusion bottlenecks.
• Drive cluster goodput by reducing GPU idle time due to data pipelines, I/O, scheduling gaps, stragglers, and recoveries.
• Build benchmarking, observability, and regression tooling; collaborate across teams to solve scale problems.

Skills

ML performance

Distributed training

GPU performance

Batch inference

Python & C++

Debugging & analytics

ML foundations

Tools

NVIDIA Triton Inference Server

TensorRT

ONNX Runtime

Ray

Nsight Systems/Compute

PyTorch Profiler

Job description

Applied Intuition, Inc. is seeking a performance engineer to optimize large-scale machine learning workloads in datacenters. You will focus on distributed training across many nodes and high-throughput inference over vast autonomy logs.

The role centers on throughput, goodput, and cost per data unit, with ownership of profiling and bottleneck resolution. The ideal candidate has hands-on ML performance engineering experience, multi-node training scale familiarity, and strong Python/C++ skills.