Tianshu Huang

I am a PhD student co-advised by Anthony Rowe and Carlee Joe-Wong at the Department of Electrical and Computer Engineering at Carnegie Mellon University. I'm interested in applying Machine Learning to Systems in ways that leverage both data-driven learning techniques and a theoretical and intuitive understanding of the properties of systems.

Previously, I obtained my B.S. in Electrical Engineering at the University of Texas at Austin in May 2021. While at UT Austin, I worked on Learning to Optimize research with Atlas Wang.

Please contact me at tianshu2@andrew.cmu.edu. You can also find my GitHub page here.

Current Research

Data-Driven Edge Orchestration for Distributed Systems
The ability to execute programs in heterogeneous distributed environments while maintaining security and performance isolation is of critical importance as edge and cloud applications become increasingly entangled. The SilverLine framework is our solution, where we propose using WebAssembly modules as an edge-friendly lighter-weight alternative to virtual machines or containers for maintaining isolation while providing portability.
In this research project, my goal is to build a statistically rigorous learning-based edge orchestration framework from the ground up integrating real-time instrumentation in our system.

Applications for Neural Radiance Fields
Since its introduction in 2020, Neural Radiance Fields (NeRFs) have garnered increasing attention. Recent advances have moved Neural Radiance Fields closer and closer to true competition with traditional techniques in computer graphics.
Applications I am interested in include problems outside the traditional NeRF "canon" (few-shot, relighting, quality, etc) such as real time reality capture and non-graphics problems, especially with assistance from non-visual sensors. In particular, I take an alternative view of NeRF research: instead of trying to best reconstruct a scene given sensor inputs, what kind of sensor inputs are necessary to reach a given target reconstruction quality, given available computational resources?

Publications

Bytecode Runtime Prediction by Matrix Factorization (In Review)
Tianshu Huang, Arjun Ramesh, Nuno Pereira, Anthony Rowe, Carlee Joe-Wong
Optimizer Amalgamation (ICLR 2022) [Poster]
Tianshu Huang, Tianlong Chen, Sijia Liu, Shiyu Chang, Lisa Amini, Zhangyang Wang

Invited Talks

Giving the Cloud an Edge with WebAssembly, WebAssembly Research Day 2022 (with Arjun Ramesh)

Students

Hanke Chen, Undergraduate (Summer 2022-) — Real Time Neural Radiance Fields
Kriti Kukreja, Masters (Fall 2022) — ML for ARENA Network Flow Understanding
Shreyashri Biswas, Masters (Fall 2022) — RL for Distributed Resource Management
Rahul Jha, Masters (Summer 2022 - Fall 2022) — SilverLine Demo and Power Measurement
Jeremy (Zhanfan) Yu, Masters (Summer 2022) — SilverLine (compilation server and file store; fault tolerance)
Anish Singhani, Undergraduate (Co-advised with Tao Jin, Spring 2022) — LIDAR-guided depth estimation
Rituraj Singh, Masters (Spring 2022) — Neural Network Training in WebAssembly

Teaching

Guest Lecture: (Duke) ECE/COMPSCI 654 Edge Computing (Federated Learning)
Guest Lecture: (CMU) 18-649 Distributed Embedded Systems (Machine Learning + Edge)
TA: (CMU) 18-661 Introduction to Machine Learning for Engineers, Spring 2022
[Lecture - Python] [Lecture - Pytorch]
TA: (UT Austin) EE 351k Probability, Statistics and Random Processes, Fall 2020 - Spring 2021

Casual Presentations

Reviewer

ICML, 2022-2023
NeurIPS, 2022
CrossFL @ MLSys, 2022