Patents:

Automatic Task Generation

Publication #20250097275

Filed: Sep. 15, 2023 - Published: Mar 20, 2025

• Designed an LLM-powered task generation system that ingests multimodal enterprise data (chat, email, voicemail, meeting transcripts) and outputs structured, actionable tasks with contextual metadata.
• Engineered a configurable AI prompt pipeline that allows admins to define task schema, enforce length/format constraints, and incorporate enterprise-specific ontologies—scaling automation to diverse workflows without retraining models.
• Reduced manual task capture by ~40 minutes per knowledge worker per day by unifying task extraction across communication channels—translating into >500 reclaimed hours per 1,000 employees daily in enterprise deployments.

Research:

Deep Neural Networks for Watermark Removal:

Final Paper written for Stanford's CS229: Machine Learning course taught under Professor Andrew Ng.

https://s3-us-west-2.amazonaws.com/secure.notion-static.com/2f020749-ea08-43c7-a0c9-1cf9fd2eec28/Deep_Neural_Networks_for_Watermark_Removal.pdf

CURIS Research: AR multi-modal interaction

Developed 2D Depth Sensing by combining Augmented Reality with the always-on nature of monocular cameras.