LeonidasĀ Tsaprounis
Senior Data Scientist
Stony Stratford, England, United Kingdom
Summary
š¤©
Rockstarš
Top SchoolLeonidas Tsaprounis is a Senior Data Scientist with six years of experience building and maintaining production-ready data science software, currently at Microsoft after senior roles at Haleon and GSK. He brings strong time series and probabilistic modeling expertise, evidenced by contributions to the popular sktime framework (adding robust forecasters and transformers) and implementing quantile functionality in TensorFlow Probability. With an academic foundation in mathematics and an MSc in Biomedical Engineering from Imperial College, he blends rigorous quantitative thinking with practical engineering. Leonidas has a track record of improving model stability and prediction interval methods, and his background includes consultancy and simulation analytics, giving him a pragmatic product-focused perspective. Based in Stony Stratford, he combines research-grade methods with hands-on open-source contributions that make complex forecasting and uncertainty estimation more reliable in production.
6 years of coding experience
5 years of employment as a software developer
MSc Biomedical Engineering with Neurotechnology, MSc Biomedical Engineering with Neurotechnology at Imperial College LondonBachelor of Science (BSc) Mathematics, Bachelor of Science (BSc) Mathematics at University of Exeter
Github Skills (17)
unit-testing10
probabilistic-programming10
python10
data-science10
scikit10
pandas10
statistics10
machine-learning10
time-series10
numpy10
forecasting10
tensorflow10
scikit-learn10
forecast10
bootstra8
Programming languages (3)
Github contributions (5)
sktime/sktime
May 2021 - Nov 2022
A unified framework for machine learning with time series
Role in this project:
Data Scientist
Data ScientistContributions:106 reviews, 20 commits, 32 PRs in 1 year 6 months
Contributions summary:Leonidas primarily contributed to the `sktime` repository, which is a time series machine learning framework, by addressing bugs related to the handling of infinite information criteria in the `AutoETS` forecaster, ensuring model stability and robustness. They introduced a `ScaledLogitTransformer` for bounded forecasting and implemented a `BaggingForecaster` using bootstrapping techniques, enhancing the framework's capabilities for prediction intervals and data augmentation. Further contributions include enhancements to the hierarchical mtype generation and fixes to indexing within the code.
forecastingtime-series-analysistime-series-regressiondata-sciencedeep-learning
tensorflow/probability
Apr 2022 - May 2022
Probabilistic reasoning and statistical analysis in TensorFlow
Role in this project:Data Scientist
Data ScientistContributions:7 commits, 1 PR, 3 comments in 25 days
Contributions summary:Leonidas primarily focused on implementing and testing the `quantile` method within the `Empirical` distribution of the TensorFlow Probability library. They added the `quantile` method to the `Empirical` distribution and included unit tests to verify its correctness across different input scenarios. Furthermore, they refactored the code by removing unnecessary checks and streamlining the import statements to align with the project's style guide. This work primarily involved the `Empirical` distribution and its associated tests.
statisticspythonprobabilistic-reasoningdata-sciencedeep-learning
Find and Hire Top DevelopersWeāve 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.