Senior Research Scientist – AI Safety

Stack

Responsibilities

• ▹Own and drive forward high-impact AI safety research themes
• ▹Contribute to the wider vision and development of Faculty's AI safety research agenda
• ▹Support Faculty's positioning as a leader in AI safety through thought leadership and stakeholder engagement
• ▹Shape the research agenda by identifying impactful opportunities and balancing scientific and practical priorities
• ▹Lead technical research within AI safety, from concept to publication
• ▹Support delivery of evaluations and red-teaming projects in high-risk domains such as CBRN and cybersecurity, with government and commercial partners

Requirements

• ▹Track record of high-impact AI research, evidenced by top-tier academic publications or equivalent experience
• ▹Proven experience in, or a clear passion for, applied AI safety — from labs, academia, or evaluation/red-teaming roles
• ▹Deep domain knowledge of language models and generative AI architectures, including fine-tuning techniques beyond API-level implementation
• ▹Practical machine learning experience with a focus on robustness, explainability, or uncertainty estimation
• ▹Proficiency with deep learning frameworks (PyTorch, TensorFlow or similar) and familiarity with the HuggingFace ecosystem
• ▹Demonstrable Python engineering experience to build and support robust research projects
• ▹Ability to conduct and oversee complex technical research projects, with excellent verbal and written communication skills

Soft skills

What we offer

• ▹Unlimited annual leave
• ▹Private healthcare and dental cover
• ▹Enhanced parental leave
• ▹Family-friendly flexibility and flexible working
• ▹Sanctus coaching
• ▹Hybrid working

About the company

Founded in 2014, Faculty is a human-centric AI company that has worked with over 350 global customers across government, finance, retail, energy, life sciences and defence to transform their performance through responsible, high-impact AI. This role sits within Faculty's Research team, which conducts red-teaming and builds evaluations for misuse capabilities in sensitive areas such as CBRN, cybersecurity and international security for leading frontier model developers and national safety institutes — work that has been featured in OpenAI's system card for o1.