Worked on the sensein/senselab repository to enhance the reliability of audio feature extraction by addressing parameter consistency in the spectrogram pipeline. Focused on correcting the use of the n_fft parameter in both spectrogram and mel-spectrogram extraction, ensuring that feature shapes remain compatible with downstream processing. Implemented automated tests in Python to verify correct parameter usage and prevent future regressions, thereby improving test coverage and model input fidelity. Applied skills in audio processing, signal processing, and testing to resolve a specific bug related to n_fft, ultimately supporting safer model training and deployment within the audio feature extraction workflow.