ML Engineer—LLM Inference & GPU Optimization (Equity)

IC Resources; San Francisco (CA); On-site; USD 200,000 - 290,000; Full time

Benefits offered by this job

401(k)

Unlimited PTO

Modern engineering workspace

Equity participation

Job summary

IC Resources seeks an engineer to accelerate production AI systems, focusing on speed and efficiency of large language model inference. You will optimize GPU-heavy pipelines and scale distributed GPU environments in a fast-moving, innovative AI company.

You will work with research and infrastructure teams to translate cutting-edge models into reliable production systems, improving latency and throughput while leveraging modern hardware.

Qualifications

• Degree in Computer Science, Electrical Engineering, or related discipline.
• Strong Python and C++ (CUDA preferred) programming skills.
• Knowledge of modern LLM serving technologies (vLLM, SGLang, PyTorch, etc.).
• Understanding of GPU architecture and parallel computing.
• Experience with model serving, distributed inference, quantization, or batching strategies.
• Strong profiling, debugging, and systems optimization skills.

Responsibilities

• Improve speed and efficiency of production AI systems.
• Evaluate performance bottlenecks and build benchmarking tools.
• Optimize inference pipelines and scale distributed GPU environments.
• Collaborate with research and infrastructure teams to deploy new modeling techniques.
• Continuously improve latency, throughput, and hardware utilization.

Skills

Python

C++

CUDA

vLLM

SGLang

PyTorch

GPU architecture

Distributed inference

Profiling

Debugging

Education

Degree in Computer Science or Electrical Engineering

Tools

CUDA toolkit

Job description

IC Resources seeks an engineer to accelerate production AI systems, focusing on speed and efficiency of large language model inference. You will optimize GPU-heavy pipelines and scale distributed GPU environments in a fast-moving, innovative AI company.

You will work with research and infrastructure teams to translate cutting-edge models into reliable production systems, improving latency and throughput while leveraging modern hardware.