Research Scientist (Control)

Location: In-person role based in the London or San Francisco office; flexible working hours and WFH arrangements are available.

Salary: £100,000–£200,000 (approximately $150,000–$270,000).

Company

hirify.global develops AI safety research and products focused on detecting and mitigating risks from loss of control, including deceptive alignment and scheming.

What you will do

• Design and run experiments to evaluate monitoring effectiveness across coding-agent failure modes and behaviors.
• Build evaluation frameworks and high-quality datasets for training and testing monitors.
• Develop and iterate monitoring approaches, balancing detection accuracy, computational efficiency, cost, and latency.
• Design agentic monitoring systems that investigate logs and identify known and novel failure modes.
• Fine-tune open-source models for high-volume production monitoring and compare them with prompted baselines.
• Incorporate red-team findings and newly discovered coding-agent vulnerabilities into monitor design and Apollo's failure-mode library.

Requirements

• 2+ years of empirical research experience with large language models or AI systems.
• Strong experience using, comparing, designing, or developing AI coding agents.
• Experience with LLM-as-a-judge systems or AI monitoring.
• Experience designing experiments, analyzing results, and iterating based on empirical findings, including prompting, scaffolding, agent design, fine-tuning, or reinforcement learning.
• Strong Python programming skills and the ability to work independently on open-ended research problems.
• On-site availability in London or San Francisco; visa sponsorship is available in both the UK and US, subject to the role and candidate.

Nice to have

• Familiarity with AI safety concepts, agent-related risks, computer security, or secure system design.
• Experience fine-tuning language models or working with smaller open-source models.
• Experience building developer tools, monitoring systems, or analyzing production logs.
• Publications or contributions to AI safety or machine learning research.

Culture & Benefits

• Flexible working hours and schedule, with WFH arrangements alongside the in-person setup.
• Equity and competitive benefits, including health, dental, and vision insurance.
• Unlimited vacation and sick leave, plus up to six months of paid parental leave.
• Retirement savings with competitive employer matching, including 401(k) for US employees.
• Provided lunch, dinner, and snacks on workdays, plus paid retreats, business trips, and relevant conferences.
• $1,000 annual professional development budget.

Hiring process

• Screening interview, followed by a three-hour take-home test, three technical interviews, and a final interview with the CEO.
• No general LeetCode-style coding interviews; AI tools may be used during the take-home test, with candidates responsible for the submitted work.