James Bergstra
Old Toronto, Ontario, Canada
Summary
🤩
RockstarJames Bergstra is a seasoned AI researcher and engineering leader with 16 years of experience building and scaling machine learning and robotics systems, currently mentoring startups as a Lab Scientist at Creative Destruction Lab. He co-founded Kindred.ai and led AI platform and robotics efforts at Ocado Technology, delivering production-grade perception, deep RL, and large-scale robotic picking systems. A prolific open-source contributor, his commits span foundational projects like Theano, Numba, scikit-learn and Hyperopt—helping improve numerical solvers, automatic differentiation, testing, and distributed hyperparameter optimization. Trained with a Ph.D. from Université de Montréal and postdoctoral work at Waterloo, he bridges theoretical neuroscience-inspired models and practical engineering, often contributing low-level performance and algorithmic improvements not obvious from product headlines. Colleagues describe him as a pragmatic scientist who moves fluid research into robust, testable code and production platforms.
16 years of coding experience
Github Skills (47)
lib10
computational-neuroscience10
restructuredtext10
computation10
numerical10
python10
ng10
image-processing10
testing10
scikit10
machine-learning10
pca10
rs10
scikit-image10
hyperopt10
Programming languages (7)
Github contributions (5)
hyperopt/hyperopt
Sep 2011 - Feb 2018
Distributed Asynchronous Hyperparameter Optimization in Python
Role in this project:
Data Scientist / ML Engineer
Data Scientist / ML EngineerContributions:711 commits, 20 PRs, 16 pushes in 6 years 6 months
Contributions summary:James made several commits focused on modifying the `hyperopt/dbn.py` file, which indicates work related to distributed asynchronous hyperparameter optimization. Their changes included adjustments to deep belief network training parameters like learning rates, l2 penalties, and epoch counts, suggesting an effort to improve model performance. Furthermore, the user added code modifications to the hyperopt/base.py file which focused on enhancements to the status of the trials, potentially assisting with debugging and result reporting.
pythonasynchronoushyperparameteroptimizationmultiprocessing
hyperopt/hyperopt-sklearn
Feb 2013 - Jul 2014
Hyper-parameter optimization for sklearn
Role in this project:ML Engineer / Data Scientist
ML Engineer / Data ScientistContributions:92 commits, 3 comments in 1 year 4 months
Contributions summary:James primarily contributed to the development of hyperparameter optimization for machine learning models within the `hyperopt-sklearn` repository. Their commits focused on implementing and refactoring an `AutoPerceptron` class, which leverages the `hyperopt` library to search the hyperparameter space of an `sklearn.linear_model.Perceptron`. The user also introduced a more general `HyperoptMetaEstimator` and made several code refactorings, further improving the framework for automated hyperparameter tuning. They also added unit tests and example demonstrations to the project, and extended the preprocessing capabilities for image data.
sklearnoptimizationhyper-parameter-optimizationmachine-learningparameter
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.