Remote CUDA Engineer — GPU Kernel Optimizer

mercor; United States; Remote; USD 110,000 - 165,000; Part time

Job summary

Mercor is seeking a CUDA Engineering Expert for a remote contract role. You will analyze and optimize GPU kernels, leveraging CUDA, HIP, and shader programming to maximize performance on modern GPUs.

The role requires 20+ hrs/week, fluent C++17, Python, and Git, plus experience with GPU programming models and profilers. You will collaborate asynchronously with AI research teams and document optimization decisions clearly.

Qualifications

• Proficient with core C++17 features and modern C++ practices.
• Strong Python scripting for tooling and automation.
• Experience with Git for version control and collaboration.
• Fluent with at least one GPU programming model (CUDA, HIP, Slang, HLSL, GLSL).
• 1+ year of GPU-focused research or professional work.
• Solid understanding of GPU profiler metrics for kernel optimization.
• Ability to optimize GPU kernels with limited contextual information.

Responsibilities

• Analyze and optimize GPU kernels for performance and hardware utilization.
• Use profiler metrics (L2 cache, throughput, occupancy) to guide improvements.
• Review kernel implementations to identify bottlenecks without deep algorithm context.
• Write and modify C++17 and Python code for GPU work.
• Apply CUDA, HIP, and shader techniques to boost performance.
• Document optimization decisions and when profiler metrics are helpful.

Skills

C++17 features

Python

Git

GPU programming models

GPU kernel optimization

GPU profiler metrics

Context-free kernel optimization

Tools

NSight Compute

Job description

Mercor is seeking a CUDA Engineering Expert for a remote contract role. You will analyze and optimize GPU kernels, leveraging CUDA, HIP, and shader programming to maximize performance on modern GPUs.

The role requires 20+ hrs/week, fluent C++17, Python, and Git, plus experience with GPU programming models and profilers. You will collaborate asynchronously with AI research teams and document optimization decisions clearly.