• Implemented a key feature within the WebRTC pipeline, enhancing cross-platform audio streaming capabilities, achieving a 15% reduction in mid-stream dropped audio connections while reducing CPU usage by 5%.

• Designed and validated a Gen AI-powered live transcription pipeline on a MacBook, reaching 91% accuracy in end-to-end tests and reducing latency by 75ms through real-time optimizations.

• Established a cross-device testbed using a phone and a laptop, validating the feasibility of end-to-end live transcription with ongoing performance improvements, resulting in stable communication with 90% transcription accuracy.