Some of my open-source contributions

Here is a non-exhaustive, continually growing list of open-source contributions (e.g. Pull Requests I had the chance to author or co-author).

Note that contributions come with many flavours and need not be technical.

scikit-learn

100+ Pull Requests, including:

• ⭐ PairwiseDistancesReduction: some innermost cornerstone implementations
• Improve the creation of KDTree and BallTree from \(\mathcal{O}(n^2)\) to \(\mathcal{O}(n)\)

400+ reviews, including 100+ reviews involving Cython, e.g:

• Common Private Loss Module with Tempita
• Scalable MiniBatchKMeans plus cln / fixes / refactoring
• Monotonic Constraints for Tree-based models
• Online implementation of non-negative matrix factorization
• Interaction constraints for HGBT
• Array API support to LinearDiscriminantAnalysis

Events:

• 🇪🇺 🐍 EuroPython 2022 Talk and Sprint
• 🇪🇺 🔬 EuroScipy 2022 Talk and Sprint
  • 📖 see the slides
• PyLadies mentoring sessions
• 🪐 JupyterCon 2023
• 🌳 Scientific Python 2023 Developer Summit
• 🇪🇺 🔬 EuroScipy 2023 Maintainer Track and Sprint

SciPy

Contributions:

• Fix KMeans++ initialisation slowness
• Faster implementation of scipy.sparse.block_diag
• Fix MLE for Nakagami nakagami_gen.fit

Reviews:

• Dinic’s algorithm for maximum_flow
• Speed up sparse.csgraph.dijkstra
• Extending _distance_pybind with additional distance metrics
• Minimum Cost Flow

Scientific Python

• Sparse Data Discussions
  • Meeting 1 presentation: Usage and needs for sparse data in scikit-learn

Miscellaneous

• My really first contribution was about pragmatic performance improvement of a C++ implementation of a fingerprint extractions algorithm.