SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Anthropic is seeking a Research Scientist/Engineer to build safety and oversight mechanisms governing how their AI models handle biological knowledge. This role sits at the intersection of applied machine learning and biosecurity, focusing on designing and running capability evaluations against frontier models, generating and curating training data for safety classifiers, and measuring their robustness against adversarial pressure in production.
Key responsibilities include: designing and executing capability evaluations to assess what new models can do in the biological domain; developing training and evaluation datasets for safety classifiers in collaboration with threat modeling experts; training and iterating on classifiers with ML engineers, optimizing for both adversarial robustness and low false-positive rates; building tooling and pipelines for fast, repeatable evaluation and classifier development; analyzing classifier performance against production traffic to identify gaps; designing and running red-teaming and stress-testing exercises; partnering with Research, Product, and Policy teams to embed biological safety throughout the model development lifecycle; contributing to external communications including model cards and policy documents; and tracking developments in biology, ML, and biosecurity for emerging risks and mitigations.
Required qualifications include strong Python proficiency and extensive scientific programming/data analysis skills; excellent grasp of ML fundamentals; deep knowledge of modern biology (high-throughput assays, functional characterization, gene synthesis, genome editing, strain construction, protein engineering); ability to build and maintain custom tooling; experience designing quantitative experiments and drawing defensible conclusions from noisy results; strong analytical and writing skills; familiarity with dual-use research concerns and biosecurity frameworks (Select Agent regulations, Biological Weapons Convention, Australia Group guidelines); comfort with ambiguity and shifting priorities; ability to work independently while collaborating cross-functionally; results-oriented mindset; and thriving in fast-paced research environments balancing rigor with rapid iteration.
Preferred qualifications include experience with large language models (prompting, fine-tuning, evaluation), training/deploying classifiers in production, developing ML methods for biological systems, adversarial robustness and red-teaming, at least 8 years of hands-on life sciences experience with deep expertise in molecular biology, drug discovery, or computational biology, and experience leading complex technical projects across multiple stakeholders.
