Overview

Broadly, I am interested in designing systems mechanisms and policies to handle trade-offs in cost, performance, and efficiency for emerging applications. Specifically, I have worked on projects related to cloud/edge resource management, big data frameworks, deep learning inference, distributed training, neural architecture search, and AR/VR. My recent work has a strong focus on improving system support for deep learning and on the practical applications of deep learning in AR/VR.

Mobile Deep Learning

My work on mobile deep learning tackles the performance issues that arise in running complex models, using cloud-based serving, and supporting secure inference. This line of research is generously supported by the National Science Foundation.

• Mobile Deep Inference: CIKM'21, SDM'21, CIKM'20, KDD'20, IC2E'20-MDInference, IC2E'20-Perseus, HotEdge'18, IC2E'18
• Model Serving: ACSOS'21, ICPE'21, Perf'20, IC2E'20-Perseus, DIDL'19, SEC'19
• Secure Inference: IC2E'21/arXiv'21-Fidelius

Mobile Augmented and Virtual Reality

With deep learning excelling in a variety of tasks that are essential to environment understanding, I am very excited to shape the future of mobile computing through algorithms and systems innovations. Ultimately, I would like to build an end-to-end framework that facilitates the AR development. I am actively looking for sponsors for this line of research. If you are interested in my work, please don't hesitate to reach out!

• Mobile AR Lighting Estimation: IMWUT'22, MM'22, IEEEVRW'22, MobiSys'21, ECCV'20, HotMobile'20, arXiv'20
• VR and Video Streaming: MMSys'20, MM'20-VVSec, MM'20-Grad

Distributed Deep Learning

My work on distributed deep learning focuses on balancing the training cost and accuracy trade-offs. I am particularly interested in providing cost-effective training solutions for deep learning practitioners via cloud-based GPU clusters. This line of work is currently supported by the National Science Foundation and was previously supported by Google Cloud. I am looking for new sponsors for continuing this research!

• Performance Characterization and Modeling: Sigmetrics'21, IC2E'21-DELI, ICDCS'20-CMDARE, ICAC'19
• Optimization: ICML'21, ICDCS'21, ICLR'22

Cloud Resource Management

Most of my early work is on resource management mechanisms and policies for traditional and distributed clouds.

• Transient Servers: ICAC'19, arXiv'19, TPDS'17, EuroSys'16, EuroSys'15, SoCC'15
• Distributed and Edge Cloud: TOIT'18, TAAS'18, IC2E'16, ICAC'15, MMSys'14, ATC'12
• Big Data Frameworks: ICDCS'20-DistStream, VLDB'19, EuroSys'16