Senior Python Engineer, LLM Evaluation

Create training and benchmarking datasets for large language models by preparing code examples, writing solutions, and evaluating AI-generated software. The work covers backend, frontend, and API development, with attention to efficiency, scalability, reliability, and performance.

The role focuses on software engineering evaluation across full-stack work, so review the duties and qualifications before applying.

• Curate code examples and provide precise solutions for model training.
• Review and refine AI-generated code in Python, JavaScript, React, Node.js, C/C++, Java, Rust, and Go.
• Assess software architecture, API design, production implementation, experiments, monitoring, and ongoing maintenance.
• Build agents that check code quality, identify recurring errors in full-stack applications, and automatically verify solutions to software engineering tasks.
• Work with cross-functional teams to improve AI coding systems against industry performance benchmarks.

What it pays and takes

• Pay: Not provided.
• Location: Remote; open to candidates in the United States.
• Schedule: Flexible, with a commitment of 20 or more hours per week.
• Engagement: Part-time contractor work, initially for one month, with potential extensions based on performance and fit.
• Skills: Strong full-stack development skills in Python and JavaScript, including React and Node.js, and experience deploying scalable, production-grade software.
• Knowledge: Software architecture, design, development, debugging, and code review, plus the ability to explain technical judgments clearly in writing and speech. English communication is required.

How it works

Apply on OpenTrain with your resume, then complete the application on the hiring site.

About AI training work

OpenTrain is a free platform where people find AI training and data-labeling projects and apply. AI training uses examples prepared and reviewed by people to teach and assess AI models, and experienced software engineers bring the technical judgment needed to evaluate generated code.