Remote job
Research Scientist - Neo
Job details
About this role
Role overview
Lead empirical research into safety risks in advanced AI models, from choosing important questions and designing experiments through analysis and publication. The work focuses on misalignment, loss of control, and harmful manipulation, including risks that may be hard to detect directly or emerge as models gain autonomy.
Responsibilities
- Define research questions and hypotheses, then plan and lead projects through to published results. - Design evaluations for risks such as evaluation awareness, sandbagging, dishonesty, sycophancy, and unsafe agent behavior. - Investigate long-horizon failures and situations where a model may conceal relevant behavior. - Assess whether alignment and evaluation methods remain effective across different frontier model ecosystems. - Partner with research engineers to build robust, reproducible experiments. - Share findings with researchers, model developers, and safety institutions.
Requirements
- A record of original research in AI safety, evaluations, or a closely related field. - Strong understanding of frontier model behavior and elicitation methods. - Ability to set a research direction and turn it into concrete published work. - Strong technical writing and close collaboration with engineers. - Comfort working in an early-stage research setting where questions and priorities can be uncertain.
Nice to have
- A PhD or comparable research output, with publications on alignment or dangerous capability evaluations. - Familiarity with frontier lab safety reports, the EU AI Act, or the NIST AI Risk Management Framework. - Mandarin reading and writing ability. - Experience in an adjacent field may transfer well, even without a specific AI safety background.
Benefits and work setup
- Based in Singapore or remote, with some working-hour overlap across regions. - Indicative salary range of $150,000–$200,000+, adjusted for location, experience, and level; higher compensation may be considered for exceptional candidates. - Competitive benefits and leave policies.