Worked on enhancing EEG data processing in the mne-tools/mne-python repository by implementing a per-channel-type average reference feature. This approach allowed the set_eeg_reference function to independently apply common average referencing (CAR) for each EEG channel type when projection was disabled and average referencing was selected, improving referencing accuracy and data integrity. The work included updating documentation and adding dedicated unit tests to ensure regression safety and maintainability. Leveraging Python, data processing, and signal processing expertise, the changes supported more reliable downstream analyses and streamlined EEG workflows, with a focus on robust testing and CI-friendly code practices throughout the development process.