At a stealth AI startup, I engineer a ~30B-parameter Diffusion Transformer that generates 2K video and synchronized stereo audio in one forward pass. I cut sampling from 50 steps to 8 or fewer using Reflow and Latent Consistency Models, and co-develop cross-modal attention to keep audio and video in sync.
I founded Sigil, an AI-native mobile app that turns spoken agreements into digital micro-contracts, with an on-device language-model voice pipeline. Previously, at CAIR, DRDO, I led 10+ engineers in deploying edge AI systems that met DRDO and Army cybersecurity and data-privacy standards with 100% compliance; in freelance work, I built RAG pipelines and machine-learning systems for financial documents, radiographic analysis, and other applications.
