Rob Zinkov is a machine learning engineer and data scientist whose work sits at the intersection of deep learning and Bayesian statistics. He focuses on how to more efficiently specify and train deep generative models, as well as how to more effectively discover a good statistical model for your data, a problem that anyone who has wrestled with priors, likelihoods, and posteriors will recognize.

Previously, Rob was a research scientist at Indiana University, where he was the lead developer of the Hakaru probabilistic programming language. His experience there gave him a practitioner’s view of the trade-offs between probabilistic programming languages and libraries, and he is a frequent guide on how tools like PyMC and Stan fit into real modeling workflows. He has spoken about practical Bayesian workflows on the DataTalksClub podcast, covering topics from MCMC and Hamiltonian Monte Carlo sampling to composing hierarchical and spatial models and diagnosing issues like multimodality.

Rob is also a strong advocate for building solid mathematical foundations, encouraging aspiring machine learning practitioners to invest in calculus, linear algebra, optimization, and statistics as the basis for a durable career in the field.