Graduate Intern - LLM Reliability and Uncertainty AI Science Assistants

Posting Title

Graduate Intern - LLM Reliability and Uncertainty for AI Science Assistants

CO - Golden

Intern (Fixed Term)

40

NLR is located at the foothills of the Rocky Mountains in Golden, Colorado is the nation's primary laboratory for energy systems research and development.

Join the National Laboratory of the Rockies (NLR), where world-class scientists, engineers, and experts are accelerating energy innovation through breakthrough research and systems integration. From our mission to our collaborative culture, NLR stands out in the research community for its commitment to an affordable and secure energy future. Spanning foundational science to applied systems engineering and analysis, we focus on solving complex challenges to deliver advanced, secure, reliable, and cost-effective energy solutions.

Our work helps strengthen U.S. industries, support job creation, and promote national economic growth.

At NLR, you'll find a mission-driven environment supported by state-of-the-art facilities, multidisciplinary research teams, and strong collaborations with industry, academia, and other national laboratories. We offer robust professional development opportunities, and a competitive benefits package designed to support your career and well-being.

The AI, Learning and Intelligent Systems group in the NLR Computational Science Center has an opening for a graduate student researcher in LLM Reliability and Uncertainty for AI Science Assistants. The researcher will investigate methods for quantifying uncertainty in LLM-based science assistants over multi-turn scientific dialogue, with an emphasis on flagging when a scientific question or task is under specified or ill-posed.

In practice, scientific questions can be vague, open-ended, or under determined. LLM-based assistants can quietly insert their own assumptions into such requests to fill the gap instead of raising concerns to their human counterpart. This internship will investigate if the assistant’s internal representations can be probed to detect these instances so they may be flagged for the user or used to trigger clarifying questions.

We are looking for a dynamic, motivated researcher with a strong technical background and an interest in AI for science, uncertainty-aware machine learning, human-AI scientific workflows, and trustworthy AI. The successful candidate must be able to work at the intersection of machine learning research and practical AI system integration.

• Research and evaluate uncertainty quantification and hallucination detection methods for multi-turn, agentic scientific workflows
• Develop probing methods that predict, from a model's internal representations, when a scientific task specification is incomplete or inconsistent and a clarifying question is warranted
• Build and instrument evaluation pipelines that capture and analyze model internal states over multi-turn scientific dialogue on HPC systems
• Conduct experiments and analyze model behavior across computational science domains and established benchmarks
• Contribute to technical documentation, research reports, publications, and presentations summarizing project progress and findings
• Develop, test, and maintain high-quality research code and evaluation pipelines

• Minimum of a 3.0 cumulative grade point average.
• Undergraduate:
  Must be enrolled as a full-time student in a bachelor’s degree program from an accredited institution.
• Post Undergraduate:
  Earned a bachelor’s degree within the past 12 months. Eligible for an internship period of up to one year.
• Graduate:
  Must be enrolled as a full-time student in a master’s degree program from an accredited institution.
• Post Graduate:
  Earned a master’s degree within the past 12 months. Eligible for an internship period of up to one year.
• Graduate + PhD:
  Completed master’s degree and enrolled as PhD student from an accredited institution.
• Must meet educational requirements prior to employment start date.
• Must meet educational requirements prior to employment start date.

• Applicants are responsible for uploading official or unofficial…