Frontier AI Safety Benchmark Architect

Space Executive; United States; On-site; USD 150,000 - 260,000; Full time

Job summary

Space Executive is seeking a top-tier AI safety researcher to define how the industry measures AI risk. You will release benchmarks every fortnight to three weeks, sometimes publicly and other times privately with labs and developers.

You own taxonomy, evaluation harness, and release quality, working with internal harm-domain experts and freelancers as needed. You will report to the CTO’s office and collaborate with the research lead on public research direction.

Qualifications

• Masters or PhD in computer science, ML, or a closely related field, or equivalent industry depth.
• At least three years designing and operating safety or security evaluations for LLMs in production.
• A track record of AI safety and security publications, with five+ papers and first authors on two or more.
• Strong engineering skills: building eval harnesses, distributed inference, vLLM, and codebase debugging.
• Ability to design a taxonomy from scratch, not just evaluate against an existing one.
• Experience leading researchers and freelancers through influence rather than formal management.
• Excellent written and verbal English with strong cross-timezone communication.
• Genuine curiosity about harms being studied and ability to learn new domains quickly.

Responsibilities

• Deliver benchmarks with a structured taxonomy, evaluation harness, and quality standards.
• Own plan and schedule, coordinate internal specialists and external experts.
• Shaping quarterly roadmap with CTO and research leads; align with commercial goals.
• Maintain industry connections; stay current with new research and conferences.
• Collaborate with frontier labs, model providers and academic partners to publish findings.

Skills

Safety evaluation design

Leadership of researchers & freelance

English communication

AI safety publishing

Taxonomy design

Engineering for eval harnesses

Education

Masters or PhD in CS/ML or equivalent

Tools

vLLM

Job description

Space Executive is seeking a top-tier AI safety researcher to define how the industry measures AI risk. You will release benchmarks every fortnight to three weeks, sometimes publicly and other times privately with labs and developers.

You own taxonomy, evaluation harness, and release quality, working with internal harm-domain experts and freelancers as needed. You will report to the CTO’s office and collaborate with the research lead on public research direction.