Official websites use .gov
A .gov website belongs to an official government organization in the United States.
Secure .gov websites use HTTPS
A lock (
) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.
Careers at CAISI
About CAISI
CAISI, within NIST at the Department of Commerce, acts as a startup within government, taking on ambitious projects to make an outsized impact.
CAISI has been designated by Secretary Howard Lutnick to serve as industry’s primary point of contact within the U.S. government to facilitate testing and collaborative research related to harnessing and securing the potential of commercial AI systems. Under President Trump’s AI Action Plan, CAISI received seventeen taskings spanning AI security research, national security evaluations, analysis of global AI competition, measurement science, interagency coordination on AI, and developing voluntary standards.
Leading AI companies partner with CAISI on a voluntary basis for support evaluating their most capable models prior to deployment. Additionally, CAISI maintains close partnerships across the federal government, including the national security community, and has built a reputation for technical excellence.
CAISI maintains offices in Washington, D.C and downtown San Francisco. Work at CAISI is in-person, however some forms of engagement may allow for other arrangements. Employment at CAISI requires U.S. citizenship.
Teams
Teams at CAISI are small and collaboration across the organization is common. Staff members and collaborators at CAISI stand out for their dedication to technical excellence, domain expertise, commitment to scientific rigor, and passion for AI. This includes, but is not limited to, software engineers, AI research engineers and scientists, cybersecurity and biological security experts, and experienced measurement scientists. Descriptions of current team activities are provided below.
Agent Security
Overview of team activities:
- Measuring and improving the security of AI systems.
- Assessing risks such as agent hijacking, data poisoning, jailbreaking, and reward hacking.
- Manual and automated red-teaming of AI agent systems.
- Developing guidelines and best practices for secure development and deployment of AI.
- Multidisciplinary measurement of AI systems in real-world settings, in application or post-deployment, through activities such as (1) field testing to gather user preferences on AI systems in real workflows and (2) uplift studies to measure AI-driven productivity across a range of tasks.
- Advancing measurement science, including (1) developing improved methods for rigorously evaluating AI system characteristics, (2) vetting measurement instruments, such as benchmarks, for statistical validity, (3) Building rigorous uncertainty estimation estimates that leverage statistical models, and (4) assessing practices for automated scoring (e.g., LLM-as-judge).
Applied Systems is hiring an AI Research Scientist
Applied Systems also participates in the NRC Research Associateship Program through the National Academies (February and August cycles)
Chem/Bio
Overview of team activities:
- Evaluating computational chemical and biological capabilities, including biomolecular prediction and design.
- Creating benchmarks to monitor advancements in performance and utility of chemical and biological AI models.
- Developing, running, and interpreting chem/bio evaluations in order to understand impacts on national security.
- Collaborating with subject matter experts within government, AI industry, and AI evaluators to improve measurement methodologies for chem/bio evaluations.
Chem/Bio is hiring an AI Evaluations Scientist, Biological AI Models
Cyber
Overview of team activities:
- Conducting automated and human-assisted evaluations of offensive cyber capabilities for vulnerability research, exploit development, and cyber workflow automation.
- Developing, running, and interpreting cyber evaluations in order to understand impacts on national security.
- Collaborating with subject matter experts within government, AI industry, and AI evaluators to improve measurement methodologies for cyber evaluations.
Cyber is hiring a Senior Cyber Offense Specialist
Frontier Assessment
Overview of team activities:
- Assessing capabilities of US and foreign AI systems and how those capabilities may evolve over time.
- Assessing general AI capabilities and their national security consequences.
- Collaborating with frontier AI labs on pre-deployment evaluations.
- Producing reports, briefings, and memos on the AI landscape for US government stakeholders.
- Building infrastructure for large-scale, quick-turnaround, high-signal evaluations.
- Designing new methodologies, e.g. for measuring partial progress on hard agent-based tasks, and for making accurate cost efficiency comparisons.
Frontier Assessment is hiring multiple Members of Technical Staff
Partnerships
Overview of team activities:
- Developing guidelines for the AI industry and government agencies, on topics including, for example, the security and robustness of AI systems and mitigating potential vulnerabilities when deploying AI agents and ensuring that AI evaluations are informative and reproducible.
- Managing CAISI’s partnership agreements and collaborations, including with frontier AI labs and academia.
- Coordinating with other teams at NIST to develop voluntary standards and engage with international standards development organizations.