This is a fully remote position, open to applicants in United States.

📋 Description

• Develop complex single-turn prompts focused on nuclear materials and safeguards, categorized as benign, dual-use, or adversarial.

• Assess AI model outputs against a specified policy standard.

• Evaluate whether the responses effectively addressed legitimate inquiries or appropriately declined dangerous questions.

• Create reference answers and clarify the technical rationale behind them.

• Provide written justifications that are comprehensible to non-experts.

• Engage in reading and writing about misuse scenarios for extended periods.

⛳️ Requirements

• Proven experience in characterizing actual nuclear material, including detection, attribution, and analytical tasks that support defensible conclusions.

• A background in nuclear forensics casework or international exercises, along with expertise in radiochemical and mass-spectrometric analysis, detection and interdiction technologies, material signature libraries, attribution methodologies, or pre- or post-detonation forensics research.

• Capability to differentiate between routine professional inquiries and those seeking harmful information.

• Excellent technical writing abilities; every decision made must be backed by a rationale that is accessible to non-specialists.

• Previous experience in technical writing, published research, or serving as an expert witness is a strong asset.

• Must refrain from using classified or export-controlled information, NDA-protected information, or data subject to prepublication review requirements.

• Obligations related to such information must be disclosed to appropriately scope the work.