Intel Corporation, 2023
PDF (internal Intel publication)

We present a multimodal, edge-based AI system for real-time PC anomaly detection with model personalization, root causing, and severity analysis. We validate this system and conduct a user study leveraging facial expression recognition to study the mapping of anomalous PC behavior to end users' frustrating experiences.

KDD-UC @ ACM SIGKDD, 2023
PDF

We exhibit the presence of racial biases in an action unit (AU) activation estimation model and highlight the non-linearity of biases across skin colors and AUs using an artifically generated dataset of faces. Further, we demonstrate the importance of dataset diversity and representation in achieving robust computer vision models.

Workshop on Verifiable Database Systems @ ACM SIGMOD, 2023
PDF

We design a benchmarking framework for standardized, transparent, and fair blockchain benchmarking on workloads derived from realistic web3 applications.

SyntheticData4ML Workshop @ NeurIPS, 2022
PDF | video

We created a dataset of synthetically generated faces as a means to explore racial biases in several publicly available computer vision facial expression recognition models.

Intel Corporation, 2022
PDF (internal Intel publication)

We develop a real-time anomaly detection system to identify and collect information about anomalous PC behavior, assisting in root-causing of frustrating or unexpected behavior and, opportunistically, manual/automatic remediation of problematic behavior toward greater user satisfaction.