Govinda Madhava

Boulder, CO

• - Built a novel low-compute spatial audio algorithm for wearables, optimizing memory and advancing user individualization.
• - Achieved perceptually identical performance while reducing compute and memory costs by 67% compared to the legacy DSP pipeline, enabling on-device spatialization by eliminating wearable → host device → wearable round-trip latency.
• - Enabled dynamic scaling of head and ear sizes, ensuring a wider variety of Apple users experience personalized spatial audio with accurate spatial and elevation localization cues.
• - Skills: Cpp, MATLAB, Python, Audio DSP, Real-Time Multi-Thread Audio Programming

Rolle, Switzerland

• - Designed an end-to-end ML pipeline delivering real-time, low-latency pitch shifting (±12 semitones) using Google Magenta DDSP and Granular Audio Synthesis, powering Tanpura Droid (1Mn+ downloads) and Swar Shala.
• - Applied time-frequency masking reducing audio bleeding in binaural datasets, improving Signal-to-Noise Ratio by 48%.
• - Optimized real-time user pitch tracking by improving Circular Buffer management & Dynamic Time Warping (DTW).
• - Skills: Python, C++, Machine Learning, Digital Signal Processing, Software Development, End-to-End Pipeline, Feature Processing

IIT Bombay, India

• - Curated a first-of-its-kind music dataset comprising 4 hrs of high quality audio, metadata & rendition-wise digitized notations.
• - Achieved 68% accuracy developing a Bi-LSTM architecture, the first auto transcription system for Indian classical vocal performances, converting audio signals to pitch contours and schematic notation.
• - Skills: Python, Machine Learning, Music Transcription, Dataset Creation

Atlanta, USA

• MS, Music Technology | Computational and Cognitive Musicology Lab (Advisor: Prof. Claire Arthur

Bengaluru, India

• B.Tech, Computer Science & Engineering