Ryan Giordano

Timeline

1. 2013

   Entering Berkeley's PhD program

   After undergraduate degrees in mathematics and engineering mechanics at the University of Illinois in Urbana-Champaign, he works at HP and teaches English and mathematics with the Peace Corps in Kazakhstan. He earns an MSc with distinction in econometrics and mathematical economics at the London School of Economics, then works on advertising experiments in Google's ads quality division. In 2013, he enters the Statistics PhD program at UC Berkeley.

2. 2015–2018

   Uncertainty estimates for variational Bayes

   In 2015, he coauthors “Linear Response Methods for Accurate Covariance Estimates from Mean Field Variational Bayes” with Tamara Broderick and Michael Jordan. Their linear response variational Bayes method estimates uncertainty in individual model variables and across variables. In 2018, the three publish “Covariances, Robustness, and Variational Bayes,” which provides improved covariance estimates and measures of sensitivity to small model changes.

3. 2019

   A Berkeley statistics doctorate

   He completes his PhD in statistics at UC Berkeley in 2019, advised by Michael Jordan, Tamara Broderick, and Jon McAuliffe. His dissertation is titled “On the Local Sensitivity of M-Estimation: Bayesian and Frequentist Applications.”

4. 2019

   The Swiss Army infinitesimal jackknife

   In 2019, he coauthors “A Swiss Army Infinitesimal Jackknife” with William Stephenson, Runjing Liu, Michael Jordan, and Tamara Broderick. The method uses a linear approximation to estimate how changing data weights affects a fitted model, reducing the need to repeatedly fit models for procedures such as cross-validation and the bootstrap.

5. 2020

   Measuring sensitivity to data removal

   With Tamara Broderick and Rachael Meager, he releases “An Automatic Finite-Sample Robustness Metric: When Can Dropping a Little Data Make a Big Difference?” in 2020. Their Approximate Maximum Influence Perturbation estimates how removing a small fraction of a sample can change conclusions. The paper finds that removing less than 1% of observations can overturn results in several influential economics papers.

6. 2023

   Professor at Berkeley

   After a postdoctoral position at MIT under Tamara Broderick, he joins UC Berkeley's Department of Statistics as an assistant professor in 2023. His research includes variational methods, Bayesian robustness, sensitivity analysis, and open software.

7. 2024

   Deterministic automatic differentiation variational inference

   In 2024, he, Martin Ingram, and Tamara Broderick publish “Black Box Variational Inference with a Deterministic Objective: Faster, More Accurate, and Even More Black Box.” Their method, deterministic ADVI, uses a fixed Monte Carlo approximation to the variational objective. It supports second-order optimization and improved covariance estimates through linear response.

8. 2026

   Survey diagnostics at Berkeley

   In June 2026, he coauthors “Locally Equivalent Weights for Multilevel Regression and Poststratification” with Alice Cima, Jared Murray, Erin Hartman, and Avi Feller. The work introduces weights that support checks of sampling variability, covariate balance, and subgroup contribution in survey models. He is an assistant professor in UC Berkeley's Department of Statistics, where he studies machine learning, variational inference, Bayesian methods, robustness quantification, and sensitivity analysis.