Gray Swan AI
Head of Biology
$230K–$280K
Remote (U.S. Only) · Remote · Director · Full-time · Posted Sep 3, 2026
- Experience
- Postdoctoral or equivalent research experience; no number of years specified.
- Education
- PhD in biology or a closely related discipline.
- Visa sponsorship
- Visa sponsorship available for exceptional candidates
About the job
Leads Gray Swan AI’s biology safety capability, evaluating biological misuse risks in frontier AI and building scalable safety systems, standards, and a specialist team.
What you'll do
- Lead adversarial evaluations of frontier AI systems for biological misuse risks across text, agentic, and multimodal capabilities.
- Work with AI researchers and engineers to develop evaluation datasets, classifiers, guardrails, benchmarks, and automated red-teaming; maintain biological risk frameworks and advise stakeholders.
- Build and lead a biology safety team, establish evaluation standards and infrastructure, and represent the company with external partners.
What they ask for
- PhD in biology or a closely related discipline, with postdoctoral or equivalent research experience.
- Deep understanding of biological misuse risks, dual-use research, pathogen biology, synthetic biology, or biosecurity.
- Experience in adversarial evaluations, biological risk assessment, AI safety research, or safety datasets for frontier AI systems.
- Familiarity with international biosecurity frameworks, dual-use governance, responsible biotechnology practices, and emerging AI safety standards.
Nice to have
- Experience developing machine learning systems, biological safety classifiers, or evaluation pipelines for high-risk biological capabilities.
- Frontier AI evaluations, red-teaming, prompt engineering, AI benchmarking, or programming/data science experience (including Python).
Pay
$230,000 - $280,000 USD base salary; $230K – $280K USD base salary
Tools
Python
AI focus
Evaluates frontier AI for biological misuse risks and develops AI safety evaluations, datasets, classifiers, guardrails, benchmarks, fine-tuned models, and automated red-teaming.
Model-written summary · the original posting is authoritative