Research Engineer​/Scientist, Agent Security

Summary

CAISI is a center of AI expertise in the U.S. government. Its Agent Security team is hiring research engineers and scientists to conduct cutting‑edge security research, perform assessments of AI vulnerabilities and robustness, and develop agent security best practices and guidelines. This notice is issued under direct‑hire authority to recruit new talent to occupations for which NIST has a severe shortage of candidates.

In this position, you will:

• Assess risks such as agent hijacking, data poisoning, jail breaking, and specification gaming through benchmark development, hands‑on red‑teaming, and analysis.
• Conduct novel research that advances the frontier of AI security measurement science.
• Collaborate closely with team members, execute complex projects end‑to‑end, and help shape team priorities.
• Design and write performant, maintainable code for research and evaluations.
• Develop guidelines and best practices for secure development and deployment of AI.
• Advise agencies across government on secure AI deployment and AI security testing.
• Work in concert with industry researchers on testing exercises and research projects.
• Clearly communicate technical concepts in memos, briefings, and research papers.

All applicants must submit a resume NOT TO EXCEED two pages on standard‑size paper. If the resume exceeds two pages, only the first two pages will be reviewed for qualifications or assessment determinations.

Basic Requirements:
Experience must be IT related; the experience may be demonstrated by paid or unpaid experience and/or completion of specific, intensive training (for example, IT certification), as appropriate. For all positions individuals must have IT‑related experience demonstrating each of the four competencies listed below. The employing agency is responsible for identifying the specific level of proficiency required for each competency at each grade level based on the requirements of the position being filled.

• Attention to Detail – Is thorough when performing work and conscientious about attending to detail.
• Customer Service – Works with clients and customers (that is, any individuals who use or receive the services or products that your work unit produces, including the general public, individuals who work in the agency, other agencies, or organizations outside the Government) to assess their needs, provide information or assistance, resolve their problems, or satisfy their expectations; knows about available products and services;
  
  is committed to providing quality products and services.
• Oral Communication – Expresses information (for example, ideas or facts) to individuals or groups effectively, taking into account the audience and nature of the information (for example, technical, sensitive, controversial); makes clear and convincing oral presentations; listens to others, attends to nonverbal cues, and responds appropriately.
• Problem Solving – Identifies problems; determines accuracy and relevance of information; uses sound judgment to generate and evaluate alternatives, and to make recommendations.

ZP‑2210‑IV:
In addition to the basic requirements, applicants must have one year (52 weeks) of specialized experience equivalent to at least the GS‑12 level (ZP‑III at NIST). Specialized experience is defined as experience demonstrating at least one of (1) expertise in building software systems that evaluate, deploy, or train AI models or systems; or (2) expertise in adversarial machine learning research or deep learning research;

or (3) expertise in assessing, studying, finding, or mitigating cybersecurity threats and vulnerabilities, especially in AI systems.

ZP‑2210‑V:
In addition to the above requirements, applicants must have one year (52 weeks) of specialized experience equivalent to at least the GS‑14 level (ZP‑IV at NIST). Specialized experience is defined as experience demonstrating at least one of (1) expertise in building state‑of‑the‑art software systems that evaluate, deploy, or train AI models or systems; or (2) advanced expertise in adversarial machine learning research or deep learning research;

or (3) advanced expertise in assessing, studying, finding, or…