Bayesian Methodologies System Dynamics Postdoctoral Researcher

Select how often (in days) to receive an alert:

We are seeking a Postdoctoral Research Associate to join the Data Science and Engineering for Nonproliferation Group in the National Security Sciences Directorate (NSSD). In this role, you will conduct fundamental research into the integration of Bayesian methodologies with system dynamics modeling, advancing statistical methods and the open-source scientific software that allow time-evolving models of complex systems to be calibrated against sparse, indirect, and uncertain observations.

The group develops and maintains an open-source Python framework for building system dynamics models and for converting those models into probabilistic programs so their parameters can be inferred from data. The successful candidate will extend the mathematical and computational foundations of that capability and help bridge the gap between Bayes theory and practical application: knowledge integration, developing robust likelihood frameworks, sampler behavior for long autoregressive time series, diagnostics and model comparison, sensitivity and identifiability analysis, intuitive model interrogation methods and calibration when observations are few or conflicting.

You will work alongside data scientists, software engineers, statisticians, and domain experts, publish and release your work openly, and apply these methods to nuclear nonproliferation and nuclear fuel cycle problems, where consequential judgments must be made from incomplete evidence.

• Collaborate with researchers and mentors to develop and apply Bayesian methods for calibrating dynamic system models and quantifying uncertainty in their predictions.
• Conduct fundamental research on the formulation of probabilistic system dynamics models, including knowledge integration, likelihood frameworks, sampler configuration and performance, convergence diagnostics, and posterior predictive checking.
• Contribute to the design, implementation, testing, and documentation of open-source scientific software that makes these methods usable and reproducible for other researchers.
• Design and execute computational studies that assess model performance, sensitivity, and parameter identifiability, and that communicate uncertainty in a form useful to analysts and decision makers.
• Deliver R&D on an ongoing basis as evidenced by publications, S&T presentations, professional community engagement, software releases, and inventions or copyrights as appropriate.
• Exercise scientific integrity in performing and communicating research.
• Ensure all work is carried out safely, securely, and in compliance with ORNL policies, standards, and procedures.
• Ability to engage in domestic and international travel as required.

• Ph.D. in statistics, applied mathematics, computer science, physics, engineering, operations research, or a related quantitative discipline.
• Demonstrated experience applying Bayesian methods to scientific or engineering problems, including prior specification, likelihood formulation, posterior sampling, and assessment of convergence and model fit.
• Proficiency in Python and the scientific computing stack, with experience developing software for scientific, statistical, or numerical computing.
• Strong written and verbal communication skills, including experience presenting scientific results to technical communities and at professional society conferences and workshops.

• Experience with probabilistic programming frameworks such as PyMC, Stan, Num Pyro, or similar.
• Experience with system dynamics or compartmental modeling—stock-and-flow formulations, feedback structure, and simulation of coupled difference or differential equations.
• Experience calibrating simulation models against sparse, indirect, aggregated, or otherwise limited observations.
• Familiarity with the computational foundations of probabilistic programming, such as automatic differentiation, tensor libraries (PyTensor, JAX), gradient-based samplers, or model transpilation and compilation.
• Experience with uncertainty quantification, global sensitivity analysis, parameter identifiability, surrogate modeling, or Bayesian model selection and comparison.
• Experience contributing to open-source scientific software, including version control, testing, continuous integration, documentation, and code review.
• Knowledge of nuclear nonproliferation, international safeguards, arms…