How to apply for Research Scientist

About Neo Research

Neo Research is a Singapore-based research and evaluations lab at the forefront of ensuring frontier AI can be safely deployed at scale. Their focus on loss-of-control and harmful manipulation risks makes them a key player in AI safety, offering a unique opportunity to shape the future of responsible AI development.

About the role

As a Research Scientist at Neo Research, you will lead cutting-edge research on loss-of-control and manipulation risks in frontier AI models, designing novel evaluation methods to detect sandbagging and deception. Your work will directly influence safety reports and publications that advance the field, with the autonomy to set research directions alongside a collaborative team.

A typical day

A typical day might involve reviewing recent frontier model releases, designing and running evaluations for sandbagging or deception, and analyzing results to draft sections of a safety report. You'll also participate in team discussions to set research priorities and collaborate on writing papers for top AI safety venues.

Who Neo Research is looking for

• Has deep expertise in AI safety research, specifically focused on frontier models and risk evaluation, with a track record of published work in top venues.
• Possesses strong machine learning and model evaluation skills, including experience with adversarial testing and red-teaming.
• Demonstrates ability to lead research direction and collaborate effectively, with experience authoring impactful safety reports or papers.
• Is actively engaged with the AI safety community, attending workshops or contributing to safety benchmarks.

Tips for this application

• Highlight your experience with evaluation awareness, sandbagging, or deception in AI models—mention specific methodologies you've developed or used.
• Tailor your research statement to explicitly address how your past work aligns with Neo Research's focus on loss-of-control and manipulation risks.
• Showcase any publications or reports you've authored on AI safety, especially those that influenced safety practices.
• Demonstrate your ability to set research direction by describing a project you led from conception to publication.
• Mention familiarity with Singapore's AI safety landscape or global collaborations, as Neo Research is based there but operates remotely.

What to cover in your cover letter

['Emphasize your expertise in frontier model evaluation and your specific contributions to understanding loss-of-control risks.', 'Describe how your research has directly addressed manipulation or deception in AI, and its impact on safety.', "Highlight your collaborative leadership style and how you've set research priorities in previous roles.", "Express alignment with Neo Research's mission to safely deploy frontier AI at scale, and your eagerness to contribute to their evaluations lab."]

Research before applying

• Read Neo Research's published safety reports and papers to understand their current evaluation methodologies and findings.
• Study their blog posts or talks on loss-of-control and manipulation risks to grasp their research philosophy.
• Familiarize yourself with Singapore's AI governance framework and how Neo Research fits into the broader safety ecosystem.
• Look into their team members' backgrounds and recent projects to identify potential collaboration areas.

Likely interview topics

Based on the job description, expect questions about:

• How would you design an evaluation to detect sandbagging in a frontier model?
• Describe a time you identified a novel risk in AI models and how you addressed it.
• How do you prioritize research directions when multiple safety risks are equally pressing?
• What are the key challenges in evaluating loss-of-control risks, and how would you overcome them?
• How do you ensure your research findings translate into actionable safety recommendations?

Common mistakes to avoid

• Don't submit a generic AI safety application—this role is specifically about frontier model evaluation, so avoid irrelevant research topics.
• Avoid downplaying the importance of evaluation awareness and sandbagging; these are core to the role, not side issues.
• Don't neglect to mention your authoring experience—Neo Research values publications and reports, so omit this at your peril.