2 months ago
Research Scientist (AI Safety)
Vacancy from Hirify Global, list of international tech companiesPlus is required to make matches and apply
Match & Cover letter
Plus required for matching with this vacancy
Job description
Text:
TL;DR
Research Scientist (AI Safety): Conducting original technical research to build foundations for safe and trustworthy AI deployment with an accent on model behavior, interpretability, steering, and reproducible experimentation. Focus on designing experiments for transformer-based LLMs, developing research infrastructure in Python and PyTorch, and translating complex findings into practical guidance for frontier labs, government bodies, and safety institutes.
Location: London, UK; hybrid working.
Company
develops and deploys responsible AI solutions for clients across government, finance, retail, energy, life sciences, and defence.
What you will do
- Design and execute technical research experiments to study and steer model behavior.
- Apply white-box and black-box methods for interpretability, steering, and behavioral analysis of transformer-based LLMs.
- Develop clean, modular, reproducible research code in Python with PyTorch and maintain supporting research infrastructure.
- Document findings in academic papers, technical reports, and presentations.
- Collaborate with senior research scientists to identify literature gaps, refine hypotheses, and define research scope.
- Translate technical insights for external frontier labs, government bodies, and AI safety institutes.
Requirements
- Strong foundation in machine learning fundamentals and transformer architectures, especially LLMs.
- Practical understanding of white-box and black-box interpretability techniques, including steering vectors or mechanistic interpretability.
- Ability to formulate hypotheses from structured technical problems.
- Strong Python and PyTorch programming skills, with experience writing clean, modular research code and using Git.
- Excellent communication skills for producing clear reports and presentations.
- Analytical rigour and exposure to AI safety areas such as robustness, evaluations, or uncertainty calibration.
Culture & Benefits
- Unlimited annual leave.
- Private healthcare and dental coverage.
- Enhanced parental leave.
- Family-friendly flexibility and flexible working.
- Sanctus coaching.
- Hybrid working environment with an inclusive, diversity-focused culture.
Be careful: if the employer asks you to log into their system using iCloud/Google, send codes/passwords, or run code/software, don't do it - these are scammers. Always click "Report" or contact support. More in guide →
