IT Specialist (AI) position at the National Institute of Standards and Technology (NIST)

The Role
This role focuses on AI Safety and Evaluation. You will be tasked with creating the scientific frameworks used to measure whether AI systems are reliable, safe, and effective in real-world environments.

Key Duties:

AI Red Teaming & Benchmarking: Develop and improve instruments for measuring AI performance, including red teaming, benchmarks, and automated scoring.

Real-World Assessment: Evaluate AI models post-deployment to identify failure modes in workplaces or sector-specific tasks.

Methodology Design: Create scientifically sound evaluation methods for end-to-end AI systems.

Communication: Translate complex technical AI concepts for non-technical policymakers and stakeholders.

Qualification Requirements
Unlike the previous Treasury roles, this position cannot be qualified for using education alone. You must demonstrate specific specialized experience.

ZP-IV (GS-12/13 equivalent):

Technical understanding of GenAI models and Machine Learning algorithms.

Experience carrying out rigorous technical evaluations to identify functionality and reliability issues.

Collaborating on AI-related projects.

ZP-V (GS-14/15 equivalent):

Advanced technical understanding of GenAI and ML.

Experience leading cross-functional AI teams.

Establishing best practices for technical standards and workflows.

Critical Application Rules (Strict)
2-Page Resume Limit: This is a hard rule. If your resume is even one sentence onto page 3, NIST will only read the first two pages. * No AI Assistance: You will be asked to certify that you did not use LLMs (like ChatGPT or Copilot) to draft your responses to the narrative questions.

Drug Testing: This position requires passing a drug test.

Security Clearance: Requires Sensitive Compartmented Information (SCI) access.

Strategic Insight: The CAISI Context
Working for CAISI at NIST puts you at the heart of the U.S. government's AI safety efforts. This is "High Stakes" AI—you aren't just building models; you are building the rules and tests that everyone else's models must pass.