Postdoctoral Fellow LLM Uncertainty Quantification

Explore LLM uncertainty quantification and calibration research as a Postdoctoral Fellow with limited hybrid work eligibility at the NRC.

This two-year fellowship develops methods for uncertainty quantification in large language model outputs and intermediate features. The work addresses confidence metrics, interpretability, trustworthy AI, and human-AI interaction. You will work autonomously under direct supervision, with guidance from the project lead, and may contribute to related LLM interpretability and evaluation research.

• Develop uncertainty quantification methods for LLM outputs
• Apply methods to intermediate model features
• Investigate certainty metrics for generated text
• Study applications in safety and evaluation
• Present research at international venues

• PhD in computer science, mathematics, statistics, physics, or related discipline
• Recent PhD or completion within six months
• Experience designing novel AI research projects
• Experience with deep learning tools and libraries
• Publication record demonstrating experimental and scientific contributions

Advance NRC research on LLM calibration, uncertainty metrics, interpretability, and trustworthy generative AI applications.