This role is for one of our clients

Compensation: $60-$90 per hour

A leading AI research organization is developing the next generation of agentic evaluation benchmarks for advanced AI models. These complex, multi-step tasks need to be reliable, unambiguous, accurately graded, and resistant to shortcuts. We are seeking experienced QA/Test Engineers to help establish the quality standards and testing processes that ensure every benchmark task measures what it is intended to measure.

Each task may represent one to two days of expert development and can span multiple technical skills. As a QA/Test Engineer, you will deeply evaluate these tasks by executing them, testing edge cases, debugging environments, and identifying potential weaknesses before they reach production evaluation. You will work closely with researchers and task authors in an iterative feedback environment.

Work Arrangement: Fully Remote — United States
Commitment: Approximately 35 hours per week
Employment Type: Full-Time

Requirements

Key Responsibilities

• Design Test Cases: Develop comprehensive test scenarios to verify that benchmark tasks function correctly, including edge cases, failure conditions, and unexpected inputs.
• Review Task Quality: Thoroughly review task descriptions, requirements, expected outcomes, and reference solutions to identify ambiguity, inconsistencies, missing requirements, and potential quality issues.
• Debug & Troubleshoot: Use Python and other development tools to investigate and resolve issues within task environments, validation logic, and automated checks.
• Develop QA Processes: Create practical, repeatable testing procedures, quality checklists, and review frameworks that can be consistently applied across benchmark tasks.
• Identify Evaluation Gaps: Examine grading logic and AI-agent execution results to detect loopholes, shortcuts, inconsistent scoring, or other factors that could compromise benchmark reliability.
• Collaborate with Researchers: Work closely with researchers and task authors to communicate issues clearly, recommend improvements, and ensure fixes are properly validated.
• Maintain Quality Standards: Help establish and enforce rigorous quality standards across complex technical evaluation datasets.

Core Qualifications

• Education: MSc or PhD in a STEM discipline, or equivalent practical experience in a research-intensive or engineering-focused environment.
• Experience: 1+ years of experience in QA, test engineering, software engineering, research engineering, or a comparable role involving significant ownership of quality.
• Proven ability to design effective test cases, develop QA processes, and investigate complex technical systems end-to-end.
• Working proficiency in Python and Git, with the ability to understand unfamiliar codebases, environments, and technical workflows.
• Strong debugging and analytical skills with an ability to systematically isolate and resolve issues.
• Exceptional attention to detail and a strong habit of maintaining clear, structured technical documentation.
• Ability to identify subtle failures, inconsistencies, edge cases, and unintended behaviors that may be overlooked by others.
• Previous experience with AI evaluation, AI training, model testing, benchmark development, or reviewing AI-generated outputs is highly desirable.
• Comfortable working independently on ambiguous and open-ended technical problems.
• Strong written communication skills and the ability to provide actionable feedback to technical stakeholders.
• A quality-focused mindset with the creativity and persistence to uncover problems others may miss.
• Ability to commit reliably to approximately 35 hours per week.

Ideal Candidate

The ideal candidate combines strong software testing and debugging capabilities with exceptional analytical judgment. You should enjoy breaking complex systems, investigating unexpected behavior, and asking whether a test genuinely proves what it claims to prove.

Experience working with AI systems, evaluation frameworks, automated grading, or complex technical benchmarks will be particularly valuable.