ML Inference and GPU Engineer
Job Description
Develop research evaluations that test reasoning about inference performance, GPU execution and model-serving systems.
What you will do
• Design demanding problems in inference optimization, CUDA kernels, quantization and serving architecture.
• Write reference solutions and scoring rubrics with the engineering reasoning behind them.
• Assess generated solutions for correctness, performance and professional judgment.
• Explain weaknesses in distributed inference, batching, caches and deployment choices where relevant to your expertise.
Required background
• One qualifying route is a relevant CS, computer/electrical engineering or related bachelor’s degree or higher, hands-on GPU/serving/quantization expertise, and systems ownership at a recognized technology or AI company.
• The research-specialist route requires 6–10 years of experience plus a doctorate, or a master’s with a strong research record.
• Current or recent practical work, clear technical explanations and independent asynchronous collaboration.
Preferred background
• SGLang, vLLM, TensorRT-LLM, Mamba/Mamba2 or IBM Granite experience.
• Inference-framework open-source contributions and distributed serving expertise, including tensor/expert parallelism, continuous batching and KV-cache optimization.
• Publications, recognized technical work and experience assessing others’ research or code.
Location and eligibility
• Remote work; country restrictions are not stated.
Availability
• Remote, flexible and asynchronous research evaluation; part-time availability can be agreed.
Pay and engagement
$105 • $179/hr
USD hourly compensation. The applicable rate and agreed work scope are confirmed before work begins.