Imri Sofer
Principal Data Scientist
Cambridge, Massachusetts, United States
Summary
👤
Senior🎓
Top SchoolImri Sofer is a Principal Data Scientist based in Cambridge, MA with 15 years of experience designing and shipping large-scale machine learning systems across e-commerce, travel, and fintech. He combines deep probabilistic modeling expertise—evidenced by contributions to the popular PyMC library improving MCMC sampling and model selection—with hands-on leadership building teams that delivered hundreds of millions in loan originations and product features that measurably increased engagement. Imri has repeatedly taken ambiguous business questions and turned them into robust production models, from real-time recommendation and sort systems at Tripadvisor and Zillow to a GenAI “Brand Voice” engine for personalized email at Klaviyo. He is fluent across the ML lifecycle: research, experimentation, productionization, and stakeholder communication, and he mentors and scales teams while keeping statistical rigor front-and-center. Notably, his academic background in cognitive science and probabilistic tooling informs a practical focus on interpretable, experiment-driven modeling.
15 years of coding experience
13 years of employment as a software developer
Computational Science, Computational Science at The Hebrew University of JerusalemDoctor of Philosophy (PhD) Cognitive Science, Doctor of Philosophy (PhD) Cognitive Science at Brown UniversityBachelor of Science (B.Sc.) Biology, Bachelor of Science (B.Sc.) Biology at Ben-Gurion University of the Negev
English, Hebrew
Github Skills (5)
mcmc10
probabilistic-programming10
python10
bayesian-inference10
variational-inference9
Programming languages (4)
Github contributions (5)
pymc-devs/pymc
Apr 2011 - Apr 2012
Bayesian Modeling and Probabilistic Programming in Python
Role in this project:
Data Scientist
Data ScientistContributions:7 commits in 11 months
Contributions summary:Imri primarily contributed to bug fixes and enhancements within the PyMC library, focusing on improving the functionality and usability of the MCMC sampling methods. They addressed issues related to the calculation of AIC and BIC, corrected errors in the MCMC sampling process, and incorporated new features like `burn_till_tuned`. Furthermore, the user updated documentation and corrected typos within the codebase. Their work demonstrates a deep understanding of Bayesian modeling and probabilistic programming principles, and skills in debugging and code maintenance.
pythonbayesian-inferencestatistical-inferencemachine-learningprobabilistic-programming
hddm-devs/HDDM-paper
May 2012 - Jul 2013
Contributions:131 commits in 1 year 1 month
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.