GPU Programming Software Engineer (Contract)

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We help freelancers discover AI training projects, build a persistent portfolio of work, and apply quickly so they can grow durable freelance careers in the AI field.

We connect people with hands-on tasks that directly shape how modern AI behaves — from annotation and transcription to model evaluation and GPU-accelerated code for model training.

About AI training work

AI training (data labeling/annotation and human feedback) is the human side of building AI systems. Contributors provide the examples, corrections, and performance tuning that modern models learn from, making this work a fast-growing, accessible path into tech.

Many roles are 100% remote, flexible, and let you fit work around other commitments. Specialist tasks that require deep technical skills—like GPU programming for model training—pay more and are at the cutting edge of AI development.

The role

You will be a contractor GPU programming software engineer focused on designing, implementing, and optimizing GPU-based software used in large language model (LLM) training tasks. The work emphasizes performance, kernel/shader tuning, and C++ host integrations.

This is an expert-level, output-driven contractor role: you’ll receive task specifications and deliver working GPU code and integrations that meet defined acceptance criteria.

• Employment: Contractor, part-time
• Time commitment: 20+ hours/week
• Location: Remote / worldwide (English required)

What you'll do

• Design GPU-based software solutions and architectures for LLM training workloads.
• Implement and optimize GPU code using CUDA, WebGPU, or GLSL as specified by the task.
• Profile and fine-tune GPU kernels and shaders to maximize throughput and efficiency.
• Develop host-side logic and integrations in C++ to support GPU execution and tooling.
• Deliver output-based task work against project specifications and acceptance criteria.

Requirements

• Advanced GPU programming experience targeting NVIDIA chips (CUDA, WebGPU, or GLSL).
• Strong proficiency in C++ for host-side development and integrations.
• Proven experience profiling and optimizing GPU kernels or shaders for performance.
• Background in graphics programming, ML acceleration, HPC, scientific computing, or another GPU-focused domain.
• Ability to produce task-complete GPU programming work that meets defined specifications.
• Expert-level experience and the capacity to move quickly into productive task work after onboarding.

Helpful background

Candidates with a portfolio or concrete project examples of GPU programming, kernel tuning, or ML acceleration work will stand out.

Comfort with fast-paced, output-driven contractor work and the ability to iterate on feedback is beneficial.

Compensation & logistics

This role pays hourly in USD on a PAY_PER_HOUR basis. The posted range is $60–$85 per hour, with a top rate of $85/hr for the most experienced contributors.

Work is remote and open worldwide; English language ability is required to participate in onboarding and task communications.

• Pay: $60–$85 per hour (USD), up to $85/hr
• Commitment: 20+ hours/week
• Employment type: Contractor, part-time
• Location: Remote — worldwide

How it works / Apply

Create a free OpenTrain account, build your profile with relevant GPU and C++ experience, and attach portfolio examples if you have them. A stronger profile helps you match to specialist tasks and demonstrate credibility across projects.

After onboarding you’ll receive task specifications, deliver outputs against acceptance criteria, and iterate with reviewers as needed. Apply through your OpenTrain profile to get started.