Prannessh KVA

Architecting the infrastructure for borderless, open-source AI. As an independent researcher, I design sub-quadratic architectures and Mamba hybrids to eliminate GPU memory ceilings, enabling massive, long-context foundation models to operate smoothly anywhere. My work deliberately crosses domains—merging hardware-aware inference, audio-to-text vector bridges, and multi-agent coordination to create out-of-the-box solutions that challenge standard attention bottlenecks. Driven by open science to democratize the next generation of machine learning.