Applied ML Research Scientist

  • Sep 2021 - Dec 2022
  • Sanas.ai
  • Leading real-time accent translation startup
  • Palo Alto, California, United States

Accent Conversion

Co-invented a real-time accent conversion network integrating TTS, ASR, and speech-to-speech translation technologies, reducing communication barriers for call center agents by 40% in controlled studies

Voice Cloning

Designed and implemented production-grade voice cloning systems using LSTM, VQ-VAE, Hu-BERT, and speaker embeddings, achieving 85% similarity scores while maintaining real-time performance

Noise Cancellation

Developed a CNN and HiFi-GAN-based noise cancellation system that improved speech recognition accuracy by 25% in high-noise environments while preserving voice quality