About the role#
The AI, Learning and Intelligent Systems group is seeking a graduate student researcher to focus on LLM reliability and uncertainty within scientific AI assistants. This role investigates how to quantify uncertainty during multi-turn scientific dialogues, specifically identifying when tasks are underspecified or ill-posed. You will explore whether internal model representations can be probed to detect these gaps, enabling the system to flag issues or request clarification from human users.
What you'll do#
- Research and evaluate uncertainty quantification and hallucination detection methods for multi-turn scientific workflows.
- Develop probing methods to predict when a scientific task specification is incomplete or inconsistent.
- Build and instrument evaluation pipelines to capture and analyze model internal states on HPC systems.
- Conduct experiments and analyze model behavior across computational science domains.
- Contribute to technical documentation and research reports summarizing project findings.
What you'll need#
- Minimum of a 3.0 cumulative grade point average.
- Familiarity with large language models, including agentic or multi-turn conversational systems.
- Experience developing or evaluating machine learning models.
- Knowledge of probabilistic machine learning or uncertainty quantification concepts.
- Hands-on experience with open-weight LLMs and modern deep learning frameworks.
- Ability to work independently and collaborate within a multidisciplinary research environment.
- Pursuing or recently completed a degree in Computer Science, Data Science, Mathematics, Statistics, Physics, or related engineering fields (Computer, Electrical, Mechanical, or Biomedical).
Location & details#
- Location: Golden, Colorado (Remote).
- Position type: Full-time, fixed-term internship with a rolling start.
- Note: This position is subject to Department of Energy access restrictions and requires the ability to obtain and maintain a federal Personal Identity Verification card.
