Senior Research Scientist, AI Safety and Biosecurity in London

You will advance our work on AI safety and biosecurity, studying the capabilities and risks of biological AI systems and developing methods to evaluate them and make them safer. We train frontier biological models on BaseData, which captures genetic diversity beyond what is available in public databases. You will work within the team that develops these models, with access to the data, training process, and biological expertise needed to study their capabilities and test safeguards.

You will build on our existing safety work and help set research priorities as our models develop. Depending on your expertise, your work could include:

• Designing evaluations and red-teaming studies that measure what biological AI systems enable in practice, including how much they add to existing tools and expertise.
• Developing threat models grounded in biology, and working with experimental scientists to establish when computational results support conclusions about real-world capabilities or risk.
• Investigating model behaviour, representations, and generalisation to understand where safety evaluations and safeguards hold up or fail.
• Developing and testing safeguards across training, model behaviour, design screening, and access, measuring both their effectiveness and their effect on useful scientific work.

We will shape the research focus together around your strengths and the problems that matter most. You will lead projects from the initial question through to experiments and methods that the team can use. Your findings will inform how we develop our models and what we publish, release, and offer through our services. You will also have scope to publish research where appropriate and collaborate with the wider AI safety and biosecurity community. You will report to the Head of AI Research and work alongside researchers in genomics, sequence modelling, and large-scale deep learning, as well as computational biologists and experimental teams across the company.

About You

You have a strong record of original research in machine learning, AI safety, computational biology, biosecurity, or a closely related field. We want to see work that has advanced understanding or changed research practice: influential papers, research tools used by others, evaluations adopted by other teams, or methods that changed how a system was built or deployed.

You bring deep expertise in a relevant area, such as model evaluation, interpretability, robustness, safeguards, biological modelling, or quantitative assessment of biological risk. You understand modern machine learning and can write the code, run the experiments, and analyse the results needed to test your ideas.

You have a PhD in a relevant field or an equivalent research record, and have led substantial research projects. You can explain your own contribution, the choices you made, and what you learned when an approach did not work.

You are curious about biology and willing to learn its details, whether you already work on biological AI or bring methods from another domain. You work closely with domain experts and recognise when a model score is insufficient evidence for a biological claim.

You exercise good judgement under uncertainty. You design experiments with appropriate controls, look for alternative explanations, and communicate findings and limitations clearly enough for others to act on them.

You work well in a small research team: sharing credit, taking criticism seriously, and helping colleagues solve problems outside your immediate project. You care about enabling useful scientific and medical applications of AI and are motivated to make them safer.

Nice to Have

Research experience in AI safety or biosecurity, including evaluations, red-teaming, or threat modelling. Experience with protein design, genomic models, structure prediction, or other biological AI systems. Experimental biology experience, or computational research validated in collaboration with an experimental team. Experience implementing and evaluating safeguards in deployed systems. Collaboration with biosecurity organisations, synthesis screening initiatives, or AI safety institutes.