AI Evaluation Specialist

Join a pioneering AI initiative focused on building next-generation evaluation benchmarks for frontier AI models. We are seeking analytical and technically skilled professionals to identify where advanced AI systems fail in subtle, real-world scenarios. Working in a red-teaming environment, you will design challenging, multi-step tasks that expose hidden vulnerabilities, reasoning gaps, and edge cases that traditional evaluations often miss.

In this role, you'll collaborate closely with AI researchers to transform discovered failure modes into high-quality benchmark tasks that improve the robustness, safety, and reasoning capabilities of state-of-the-art AI systems.

This is a fully remote, full-time engagement requiring approximately 35 hours per week .

Requirements

Key Responsibilities

• Investigate how frontier AI models perform across coding, machine learning, analytical reasoning, and complex problem-solving tasks.
• Identify hidden failure modes, edge cases, reasoning errors, and vulnerabilities that may not be apparent through standard testing.
• Design challenging evaluation tasks that accurately measure AI capabilities while remaining objective and reproducible.
• Document findings with clear technical explanations, supporting evidence, and reproducible methodologies.
• Collaborate with benchmark designers and AI researchers to refine evaluation tasks, eliminate loopholes, and strengthen grading criteria.
• Share insights and recommendations with cross-functional teams to continuously improve AI evaluation quality and benchmark coverage.

Required Qualifications

• Master's degree, PhD, or equivalent practical experience in a STEM discipline involving research, coding, or advanced data analysis.
• Minimum 1 year of experience in AI research, research engineering, security research, AI evaluation, or a related technical field.
• Demonstrated experience identifying vulnerabilities, adversarial behaviors, edge cases, or failure modes in Large Language Models or other machine learning systems.
• Strong proficiency in Python and Git , with the ability to build custom scripts for experimentation, testing, and analysis.
• Solid understanding of modern Large Language Models, their strengths, limitations, and evaluation methodologies.
• Experience with AI benchmarking, model evaluation, adversarial testing, prompt engineering, or dataset creation is highly desirable.
• Excellent analytical thinking, creativity, and attention to detail, with the ability to solve ambiguous, open-ended problems independently.
• Outstanding written communication skills for documenting technical findings clearly and accurately.
• Ability to commit approximately 35 hours per week on a consistent basis.

Preferred Qualifications

• Experience with AI safety, red teaming, adversarial machine learning, or security research.
• Background in benchmark design, evaluation framework development, or AI quality assurance.
• Experience creating reproducible technical experiments and documenting complex failure analyses.
• Familiarity with frontier AI research methodologies and model capability assessments.

Why Join

• Help shape the future of AI evaluation by identifying critical weaknesses before they reach production.
• Work on cutting-edge AI systems alongside researchers developing next-generation language models.
• Apply your technical expertise to improve AI reliability, reasoning, and robustness.
• Contribute directly to benchmark development that influences the evolution of advanced AI technologies.
• Enjoy the flexibility of a fully remote engagement while working on impactful research initiatives.

Equal Opportunity

We are committed to fostering an inclusive and diverse environment where all qualified applicants receive equal consideration. Reasonable accommodations are available throughout the application and engagement process.

Contract & Engagement Details

• Independent contractor engagement.
• Fully remote with flexible working hours.
• Expected commitment of approximately 35 hours per week .
• Project duration may be extended, shortened, or concluded based on project requirements and individual performance.
• Work does not require access to confidential or proprietary information from any current or former employer.
• Payments are issued weekly based on approved work completed.
• At this time, we are unable to support H1-B or STEM OPT candidates.