Research Scientist (Control)

Apolloresearch; Greater London; On-site; GBP 100,000 - 200,000; Full time

Benefits offered by this job

Market-competitive salary

Equity

Flexible hours

Unlimited vacation

Unlimited sick leave

Parental leave (up to 6 months)

Health, dental, vision insurance

Employer matching 401(k) or equivalent

Lunch, dinner, snacks

Paid work trips & conferences

Yearly professional development budget

Relocation support & visa fees

Job summary

Apollo Research in London or San Francisco is seeking a Research Scientist (Control) to join the AGI safety product team and help transform AI control research into practical safety tools. You will work closely with the CEO and product engineers on Watcher, a monitoring system for coding agents, shaping research directions and helping scale safety at speed.

You will enjoy empirically-driven work that translates AI risks into concrete detection methods, iterates quickly based on data, and makes a

Qualifications

• Experience applying empirical research to AI safety.
• Ability to design experiments to test monitor effectiveness.
• Familiar with red-teaming and adversarial evaluation.
• Experience with monitoring systems and prompts.

Responsibilities

• Research & Development: systematically collect and catalog coding agent failure modes from various sources.
• Design and conduct experiments to test monitor effectiveness across failure modes and agent behaviors.
• Build and maintain evaluation frameworks to measure progress on monitoring capabilities.
• Build and maintain high-quality datasets to train and test monitors.
• Iterate on monitoring approaches based on results, balancing detection accuracy with cost.
• Stay current with AI safety research, agent failures, and detection methodologies.
• Stay current with research into coding security and safety vulnerabilities.
• Monitor Design & Optimization: develop a library of monitoring prompts for failure modes.
• Experiment with reasoning strategies and outputs to improve reliability.
• Design and test hierarchical monitoring architectures and ensembles.
• Optimize log pre-processing pipelines to extract signals efficiently.
• Implement scaffolding approaches for monitors, including chain-of-thought and multi-step verification.
• Fine-tuning & Red-teaming: fine-tune open-source models for production monitors.
• Design and build agentic monitoring systems to investigate logs and identify failure modes.
• Build automated red-teaming pipelines that attack monitors at scale.
• Design iterative adversarial games where red and blue teams continuously improve.

Skills

Research

AI Safety

Experiment design

Monitoring design

Red-teaming

Tools

Open-source models

Evaluation frameworks

Job description

Opportunity

Join our AGI safety product team to transform AI control research into practical tools that reduce risks from AI. As a Research Scientist (Control), you will work closely with the CEO, control researchers, and product engineers. We are building Watcher, a monitoring tool for coding agents, and our monitoring research agenda translates compute into safety at scale. You will join a small team with significant opportunity to shape the team and technology and to take on responsibility quickly.

You will like this opportunity if you are passionate about using empirical research to make AI systems safer in practice, enjoy translating theoretical AI risks into concrete detection mechanisms, thrive on rapid iteration and learning from data, and want your research to directly impact real-world AI safety.

Responsibilities

• Research & Development

• Systematically collect and catalog coding agent failure modes from real-world instances, internal deployments, public examples, research literature, and theoretical predictions
• Design and conduct experiments to test monitor effectiveness across different failure modes and agent behaviors
• Build and maintain evaluation frameworks to measure progress on monitoring capabilities
• Build and maintain high-quality datasets to train and test monitors
• Iterate on monitoring approaches based on empirical results, balancing detection accuracy with computational efficiency
• Stay current with AI safety research, agent failures, and detection methodologies
• Stay current with research into coding security and safety vulnerabilities

• Monitor Design & Optimization

• Develop and maintain a comprehensive library of monitoring prompts tailored to specific failure modes (e.g., security vulnerabilities, goal misalignment, deceptive behaviors)
• Experiment with reasoning strategies and output formats to improve monitor reliability
• Design and test hierarchical monitoring architectures and ensemble approaches
• Optimize log pre-processing pipelines to extract signals while minimizing latency and costs
• Implement and evaluate scaffolding approaches for monitors, including chain-of-thought reasoning, structured outputs, and multi-step verification

• Fine-tuning & Red-teaming

• Fine-tune open-source models to create efficient monitors for high-volume production environments
• Design and build agentic monitoring systems that autonomously investigate logs to identify known and novel failure modes
• Build automated red-teaming pipelines that attack monitors at scale
• Design iterative adversarial games where a red team and blue team continuously attack and defend

• Role perks

• This role offers a market competitive salary, equity, and competitive benefits.
• Salary: 100k - 200 GBP (~150k - 270k USD)
• Flexible work hours and schedule
• Unlimited vacation and unlimited sick leave
• Up to 6 months of paid parental leave
• Comprehensive health, dental, and vision insurance
• Retirement savings with employer matching
• Lunch, dinner, and snacks provided on workdays
• Paid work trips, staff retreats, and relevant conferences
• Yearly professional development budget
• Relocation support and visa fees (if applicable)
• Time Allocation: Full-time
• Location: In-person role in London or San Francisco with flexible hours and work-from-home arrangements
• Visa sponsorship: We sponsor visas in the UK and US; sponsorship is not guaranteed for every role or candidate, but we will work to find the right visa route if an offer is made

About the Team

The Product team is new. You will work closely with Marius Hobbhahn (CEO, leads the monitoring team), Victor Gillioz (Research Scientist), Monika Jotautaitė (Research Scientist), and our product engineers. You will interact with other SWEs and researchers as we aim to be our own customer by using our products internally for research. You can find the full team information here.

About Apollo Research

The rapid rise in AI capabilities offers opportunities but also risks from Loss of Control. We focus on detection and mitigations of scheming and related risks and work with frontier AI companies to test models before deployment. We aim for a culture of truth-seeking, goal orientation, constructive feedback, and friendliness. More details about working at Apollo can be found here. We are developing tools like Watcher to prevent harms from widely deployed AI systems.

Equality Statement: Apollo Research is an Equal Opportunity Employer. We value diversity and provide equal opportunities regardless of age, disability, gender identity, marriage and civil partnership, pregnancy and maternity, race, religion or belief, sex, or sexual orientation.

How to Apply

Please complete the application form with your CV. A cover letter is not required. Feel free to share links to relevant work samples.

Interview process: multi-stage with a screening interview, a take-home test (about 2 hours), three technical interviews, and a final interview with the CEO. Technical interviews relate to tasks you would perform on the job. If you want to prepare, consider building simple monitors for coding agents and testing them on your own Claude Code, Cursor, or Codex traffic.

Your Privacy and Fairness: We protect your data, uphold fairness, and use AI-powered tools to assist with tasks such as resume screening in compliance with AI governance frameworks. Resumes are screened by humans and final decisions are made by our team. For questions about data processing or concerns about fairness, contact us at [e-mail].