Principal GPU Inference & Workload Optimizer

STN, Inc.; San Francisco (CA); On-site; USD 200,000 - 320,000; Full time

Job summary

Horizon AI is seeking a Principal Inference & Workload Optimization Lead to own how customer AI workloads run on committed capacity and to build autonomous optimization that scales beyond a single expert.

You will profile workloads, establish baselines on current NVIDIA hardware, develop agent-executable improvements, and partner with customer and internal teams to optimize GPU allocation and routing. This role emphasizes measurable performance and scalable impact across clusters.

Qualifications

• Deep GPU performance engineering across workloads.
• Experience profiling memory bandwidth and kernel efficiency.
• Ability to translate findings into agent-executable optimizations.

Responsibilities

• Profile customer training and inference workloads to identify bottlenecks.
• Establish baselines and define good performance per hardware generation.
• Convert expertise into agent-executable optimizations applied across clusters.
• Collaborate with operating teams on workload tuning and capacity decisions.
• Assume secondary responsibility for internal model serving with cost and latency awareness.

Skills

GPU performance engineering

Profiling tooling

Customer communication

Tools

Nsight

DCGM

TensorRT-LLM

NCCL

Job description

Horizon AI is seeking a Principal Inference & Workload Optimization Lead to own how customer AI workloads run on committed capacity and to build autonomous optimization that scales beyond a single expert.

You will profile workloads, establish baselines on current NVIDIA hardware, develop agent-executable improvements, and partner with customer and internal teams to optimize GPU allocation and routing. This role emphasizes measurable performance and scalable impact across clusters.