Research

Research Statement

I design resource-efficient machine learning for health sensing—methods that run reliably on wearables and ambient devices with limited compute and power. My work spans eating-behavior understanding, wrist-worn energy-expenditure modeling, and continuous stress monitoring, with an emphasis on end-to-end systems that are accurate, interpretable, and ready for real-world deployment.

Ongoing Projects

kCal Estimation Pipeline

An end-to-end pipeline from images/videos: food classification → segmentation → monocular depth → volume → nutrition estimation. Designed for reproducibility and pluggable models.

Wi-Fi Indoor Positioning (RSSI + FTM)

Probabilistic fusion of RSSI fingerprinting and FTM ranging with LOS/NLOS analysis and cross-site generalization. Targeting robust, site-independent localization.