Worked on stabilizing real-time audio diarization in the WhisperLiveKit repository by addressing a critical bug in PCM data handling. Focused on improving the reliability and accuracy of speaker segmentation when diarization is enabled, the solution involved refining the logic for adding PCM audio arrays to the diarization queue. This adjustment reduced processing errors and ensured correct sequencing of audio frames across sessions. The work was implemented using Python, leveraging asynchronous programming techniques and audio processing expertise. The fix was thoroughly reviewed and documented to align with the existing architecture, resulting in a more robust end-to-end diarization flow without data loss.