Remote | GPU Programming Software Engineer — $60–$95/hour at 24-MAG in New York, New York, United States

Job description

• We are sharing a specialised consulting opportunity for experienced GPU Programming Software Engineers with strong expertise in CUDA, WebGPU, GLSL, C++, GPU architecture, kernel and shader optimisation, and high-performance parallel computing to contribute to an advanced AI training and GPU-programming evaluation project.
• Selected professionals will design and implement GPU-focused technical tasks, optimise kernels and shaders, analyse GPU performance, and review AI-generated solutions for correctness and efficiency. The work is suited to engineers with practical GPU-programming experience across graphics, machine-learning acceleration, scientific computing, HPC, or comparable GPU-intensive domains.

Responsibilities

• GPU Development & Optimisation
• Design and implement GPU workloads using CUDA, WebGPU, GLSL, or comparable technologies
• Develop and optimise CUDA kernels and shader-based workloads
• Apply appropriate parallelisation strategies and GPU execution models
• Improve memory-access patterns, throughput, latency, occupancy, and resource utilisation
• Ensure performance improvements preserve correctness
• Performance Profiling & Architecture
• Profile GPU applications to identify computational and memory bottlenecks
• Analyse thread organisation, synchronisation, execution behaviour, and architecture-specific constraints
• Compare alternative GPU implementations and optimisation approaches
• Identify inefficient or incorrect GPU execution patterns
• Document performance findings and technical trade-offs clearly
• C++ Integration & GPU Systems
• Develop host-side logic and integrations in C++
• Manage CPU–GPU communication, data transfer, and execution workflows
• Integrate GPU functionality into broader software systems
• Structure GPU workloads within maintainable application code
• Debug performance and correctness issues across host and device components
• AI Task Development & Evaluation
• Design technically rigorous GPU-programming tasks for AI training
• Create problems testing GPU architecture, optimisation, and performance reasoning
• Define clear expected outcomes and evaluation criteria
• Review AI-generated GPU solutions for correctness, efficiency, and scalability
• Provide structured technical feedback supporting model improvement

Requirements

• Advanced professional experience with GPU programming
• Strong proficiency with CUDA, WebGPU, GLSL, or another GPU technology capable of targeting NVIDIA hardware
• Strong C++ proficiency
• Deep understanding of GPU architecture and parallel execution
• Experience profiling and optimising GPU kernels or shaders
• Strong performance-engineering and debugging skills
• Background in graphics, machine-learning acceleration, scientific computing, HPC, or another GPU-intensive domain
• Experience analysing memory-access patterns and GPU resource utilisation
• Ability to reason about architecture and performance trade-offs
• Strong technical problem-solving and communication skills
• Experience creating or reviewing rigorous programming problems is advantageous
• No prior AI-training or model-evaluation experience is required

Skills

Required

• CUDA
• WebGPU
• GLSL
• C++
• GPU architecture
• Kernel optimisation
• Shader optimisation
• High-performance parallel computing
• Performance profiling
• Debugging
• Memory-access pattern analysis
• GPU resource utilisation