Research Engineer

About Neo Research

Neo Research operates as an independent organization located in Singapore, specializing in advanced safety research and evaluations. Their work centers on identifying and mitigating risks associated with loss-of-control and harmful manipulation in cutting-edge AI models.

Key Responsibilities

• Develop and execute safety assessments specifically targeting frontier AI models with an emphasis on preventing loss-of-control and harmful manipulation scenarios.
• Create and sustain frameworks for AI agents, integrate various tools, and manage evaluation infrastructure effectively.
• Ensure reproducibility in experiments by managing sampling processes, detailed logging, maintaining controlled environments, and analyzing outcomes thoroughly.
• Interpret scientific research inquiries and translate them into rigorous, executable evaluation protocols.
• Contribute to authoring the methodology and infrastructure documentation sections in published safety research reports.

Required Qualifications and Skills

• Proficient in Python programming language.
• Direct experience conducting or developing safety evaluations for large language models (LLMs).
• Expertise in agent-based evaluation frameworks, such as Inspect or comparable systems.
• Strong software engineering principles including testing, version control, and working with containerized environments.
• Meticulous understanding of elicitation techniques including scaffolding, tool accessibility, sampling strategies, and recognizing evaluation failures.
• Excellent ability in technical writing to clearly document findings and methodologies.

Preferable Additional Experience

• Background in assessing dangerous AI capabilities.
• Familiarity with infrastructure for sandboxing agents using technologies like Docker and Kubernetes.
• Knowledge of frontier lab safety report publications.
• Ability to read and write Mandarin.

Employment Details

• This is a full-time position based in Singapore or available for remote work.