Post-Doctoral Fellow in AI Uncertainty Quantification and Conformal Prediction The George Washington University in United States

George Washington University, via the NIST Professional Research Experience Program (PREP), is hiring a Post-Doctoral Fellow for AI uncertainty quantification and conformal prediction in Gaithersburg, Maryland at the NIST campus.

The project focuses on making black-box AI systems more mathematically honest under shifted or out-of-distribution data. The fellow will build a Python-based numerical testbed for complex sequential decisions, integrate distribution-free statistical theory into deep learning architectures, and develop external monitoring layers that can detect overconfidence and flag unsafe inputs before failures occur.

Research topics include uncertainty quantification, conformal prediction, calibration, multivariate time-series monitoring, Bayesian loss analysis, latent-layer auditing, and simulation-based uncertainty propagation. The work also involves runtime inspection of PyTorch/TensorFlow models, tensor manipulation, and designing decision thresholds for autonomous versus human-escalated actions.

Eligibility highlights: a Ph.D. in Statistics or a closely related highly quantitative field; deep expertise in conformal prediction, multivariate time series, and hierarchical mixed-effects modeling; strong Python skills; and hands-on deep learning experience with PyTorch or TensorFlow. Experience building and stress-testing agentic AI systems is also requested.

Funding: the post is described as a post-doctoral fellow position through NIST PREP, but no stipend amount or detailed funding package is provided in the post.

Application: the post includes a short link to the announcement and a LinkedIn update link, but no explicit deadline is stated.