Hi, I'm Darius!
Current Role
I am a Research Scientist at Netflix, working on speech and audio as part of the Media Production Algo team. Previously, I was an Applied Scientist at Amazon AGI on the Audio Foundation Model team, working on audio understanding and reasoning.
Recent Highlights
-
Neural audio coding for high-resolution music at the University of Rochester
I gave an invited talk on neural audio coding for high-resolution music applications at the Audio Information Research (AIR) Lab, University of Rochester, hosted by Dr. Anastasia Kuznetsova.
-
HE-NAC accepted for publication in IEEE TASLP
HE-NAC: High-Efficiency Neural Audio Coding via Hierarchical and Residual Multi-Band Quantization improves compression efficiency and audio quality through a hierarchical, residual approach. The published paper and code are now available.
-
Joined Netflix as a Research Scientist
I work on generative AI for speech and audio as part of Netflix's Media Production Algo team.
-
Earned my PhD from Indiana University
My dissertation, Efficient Native Neural Sub-Band Coding Through Residual Feature Representation Within Hyper-Autoencoded Reconstruction Propagation Networks, is available through ProQuest. It explores high-fidelity neural audio compression through HARP-Net, Multi-band HARP-Net, and HE-NAC, combining hierarchical representations, subband bitrate allocation, and residual learning to preserve perceptually important audio detail.
Research Focus
My research focuses on audio and speech source separation and neural audio coding, with additional work in audio classification, automatic speech recognition, and generative modeling. I have published in venues such as IEEE Transactions, ICASSP, WASPAA, and ISMIR.
Background
I earned my PhD in Intelligent Systems Engineering from Indiana University, advised by Dr. Minje Kim.
Before my PhD, I worked as a full-time engineer on the Apple Music Apps Team (Logic Pro, GarageBand). During my PhD, I completed research internships at:
