Remote job
Research Engineer - Neo
Job details
About this role
Role overview Build experimental systems for research into safety risks from increasingly capable language models, including misalignment, loss of control, and harmful manipulation. You’ll partner with research scientists to turn open questions into reliable evaluations, own substantial parts of the research infrastructure, and contribute technical methods and analysis to published work.
Responsibilities - Design, implement, and run evaluations of model misalignment and loss-of-control risks. - Build agent environments, scaffolds, and tool integrations for studying behavior under greater autonomy. - Develop infrastructure for long-running experiments involving many model calls, actions, tools, and environment states. - Ensure experiments are reproducible through reliable model access, configuration, environment management, logging, and analysis. - Work with researchers to shape tractable experiments and distinguish model behavior from evaluation artifacts. - Create tools to inspect and compare large collections of model trajectories across models and conditions. - Contribute experimental methodology, infrastructure, and technical analysis to research publications.
Requirements - Strong Python and general software engineering skills. - Experience with language models, agent systems, or model evaluation. - Ability to turn an open-ended research question into a dependable experimental setup. - Strong practices around reproducibility, testing, observability, and data integrity. - Skill investigating unexpected results across both models and supporting infrastructure. - Clear technical writing and communication, and comfort with evolving priorities in an early-stage setting.
Nice to have - Background in AI safety or dangerous-capability evaluation, evaluation frameworks, agent scaffolds, or tool-use environments. - Experience with distributed inference, large-scale API experiments, Docker, Kubernetes, or analysis of model trajectories. - Familiarity with frontier-model safety reports or Mandarin.
Benefits and work setup - Based in Singapore or remote, with some working-hour overlap across regions expected for remote colleagues. - Indicative compensation is $120,000–$180,000+, adjusted for location, experience, and level; competitive benefits and leave policies are offered.