I built an MCP- and RAG-based document compliance agent that matches multimodal documents and validates them against company requirements. Its use cases include reviewing loan applications against lending criteria and checking insurance claims against policy requirements.
At Simon Fraser University, I researched neural shape analysis and language-guided scene understanding. My paper evaluating neural shape analysis methods was accepted for an oral presentation at CRV 2023.
At Huawei, I worked on single-view scene reconstruction using deep learning and Depth Transformer (DPT) for depth estimation. Earlier, at Adobe, I worked as a Research Intern on a vision-and-language model for visual question answering.
My projects include a local, privacy-focused document content matcher with a FastAPI backend and an AWS data pipeline for recommendation-system training data. I’ve also published work on trajectory representations, 3D shape analysis, and 3D indoor scenes.
