How to apply for Research Scientist - Neo

About SASH

Neo Research is an independent AI safety research organization based in Singapore, focused on frontier risks in increasingly capable AI systems. It specifically studies the open-weight model ecosystem and the growing frontier-model ecosystem in Asia. The parent organization, SASH, is a Singapore-based nonprofit that builds AI safety research capacity and governance expertise across Asia.

About the role

This is a research scientist role where you identify important questions and own research projects end-to-end, from initial hypothesis and experimental design through to analysis and publication. The work focuses on misalignment, loss of control, and harmful manipulation, including whether models pursue unintended objectives, conceal problematic behavior, recognize and adapt to evaluations, evade oversight, or become less safe with greater autonomy. The role is remote and full-time.

A typical day

A typical day likely involves reading recent work on frontier AI safety, designing or running experiments on model behavior, analyzing results, and writing up findings for publication. You would also communicate with a small remote team and possibly external collaborators. The exact routine is not specified, so ask about daily workflows, meeting cadence, and engineering support during interviews.

Who SASH is looking for

• Has experience designing and running empirical research projects end-to-end, from hypothesis through publication.
• Understands frontier AI safety concepts such as misalignment, loss of control, harmful manipulation, evaluation awareness, and oversight evasion.
• Is familiar with open-weight models and the Asian frontier-model ecosystem, or is willing to study them closely.
• Can work independently in a small organization and materially shape the research agenda.

Tips for this application

• Read Neo Research's published work and current projects before applying, and reference specific papers or research directions in your application.
• Show that you can identify important but difficult-to-measure risks, since the job description emphasizes producing rigorous empirical evidence about such risks.
• Describe at least one research project you owned end-to-end, including the hypothesis, experimental design, analysis, and publication outcome.
• Explain why you want to focus on open-weight models and the Asian frontier-model ecosystem, not just AI safety in general.
• Since the role is remote, state how you manage independent research work and communicate progress without close supervision.

What to cover in your cover letter

['Your track record of owning empirical research projects from hypothesis to publication.', 'Specific knowledge of frontier safety risks such as misalignment, loss of control, harmful manipulation, evaluation awareness, or oversight evasion.', 'Your interest in open-weight models and the Asian frontier-model ecosystem.', 'How you would use the freedom at a small organization to shape the research agenda.']

Research before applying

• Read Neo Research's published papers and current project descriptions to understand their specific research questions and methods.
• Study the open-weight model ecosystem in Asia, including which organizations release models and what safety evaluations exist.
• Look into SASH's work on AI safety research capacity and governance expertise across Asia to understand the broader context.
• Find out what engineering support and external relationships Neo Research offers, since the job description mentions these help turn ideas into serious empirical work.

Likely interview topics

Based on the job description, expect questions about:

• How you would design an experiment to measure whether a model conceals problematic behavior or recognizes evaluations.
• Your view on what frontier safety risks are most understudied in open-weight models developed in Asia.
• A detailed walkthrough of a research project you owned end-to-end, including what you would do differently.
• How you prioritize research questions when many important risks are difficult to measure.
• How you would work independently and communicate with a small remote team.

Common mistakes to avoid

• Applying with only general AI safety interest and no specific empirical research experience.
• Ignoring the focus on open-weight models and the Asian frontier-model ecosystem in your application.
• Failing to show you can own a project end-to-end, for example by only describing team-based or delegated work.