Worked on the livekit/agents repository to enhance the stability of the text-to-speech (TTS) pipeline by addressing a key audio processing issue. Focused on backend development using Python, the work involved correcting the output MIME type in the Sarvam TTS plugin from audio/mp3 to audio/wav. This change ensured that audio output aligned with downstream processing requirements, reducing errors and improving compatibility across the pipeline. The update was fully traceable through a dedicated commit and issue reference, reflecting a methodical approach to bug resolution. The contribution emphasized reliability and maintainability in audio processing workflows, leveraging expertise in Python and backend systems.