Skip to content

Tianshu Huang / Research

My past and current work spans a wide range of topics in machine learning, including large transformer models (ICCV '25), statistical learning (MLSys '25), NeRF-style neural-implicit inverse rendering (CVPR '24), and meta-learning (ICLR '22). I also actively support systems researchers by providing machine learning, statistics, and data science expertise (OOPSLA '25, RTAS '25, EuroSys '25), while also supporting machine learning researchers working with real-world sensors and systems.

Currently, I'm focused on rapidly scaling and commercializing my recent work on creating foundational models for radar via a technology transfer collaboration with Robert Bosch GmbH, a leading automotive radar manufacturer.

PhD Thesis: Learning on Spectrum for Radar-Enabled 3D Perception

3D perception systems should use learning-based methods on unfiltered 4D spectra. When fused with cameras and trained at scale, spectrum-based systems will far outperform classical signal processing-based methods, and match the quality of lidar-based systems even when using only low-cost single-chip radars.

Committee: Anthony Rowe, Carlee Joe-Wong, Deva Ramanan, Zico Kolter

Machine Learning for Radar

Radars are an ideal complement to cameras for applications such as autonomous driving: both are inexpensive, solid-state sensors, with cameras boasting fine angular resolution and radars providing depth resolution and robustness to adverse conditions. Unfortunately, unlike visual images or Lidar points, radar data are harder to interpret, and lack a large body of existing research. In this project, my goal is to develop machine learning-based methods to interpret radar data both spatially and semantically, potentially replacing Lidar as the primary means of 3D perception in robotics and beyond.

Radar Imaging

Learning for Distributed Systems

Recent advances in lightweight, bytecode-based virtualization — WebAssembly — raise the possibility of flexibly executing distributed programs in heterogeneous environments. While this promises substantial opportunity for optimizing over static, homogenous deployments, exploiting this opportunity requires mastering key building blocks for managing distributed systems. Along with my collaborators, I explore key concerns including orchestration, performance analysis, as well as debugging and anomaly detection. My approach emphasizes going beyond black-box approaches (in both a statistical and a systems sense) using techniques like instrumentation injection and statistical machine learning approaches.

Silverline

Publications

Towards Foundational Models for Single-Chip Radar
Tianshu Huang, Akarsh Prabhakara, Chuhan Chen, Jay Karhade, Deva Ramanan, Matthew O'Toole, Anthony Rowe
ICCV 2025

We train a foundational model for single-chip mmWave radar. Details to follow!

Unveiling Heisenbugs with Diversified Execution
Arjun Ramesh, Tianshu Huang, Jaspreet Singh Riar, Ben Titzer, Anthony Rowe
MLSys 2025 | Artifact Available

Heisenbugs - bugs which change their behavior under observation - are among the toughest challenges when debugging programs. We propose to catch these bugs using stochastic instrumentation, and demonstrate the surprising effectiveness of debugging across a diverse set of platforms.

[Github][Instrumentation Framework]

Unveiling Heisenbugs with Diversified Execution

Interference-Aware Edge Runtime Prediction with Conformal Matrix Completion
Tianshu Huang, Arjun Ramesh, Emily Ruppel, Nuno Pereira, Anthony Rowe, Carlee Joe-Wong
MLSys 2025 | Artifact Available — Artifact Functional

Performance prediction in heterogeneous systems such as edge computing is best formulated as matrix completion, which can be extended to handle complex, edge-specific concerns such as interference and uncertainty quantification.

[Github][Zenodo][Slides]

Interference-Aware Edge Runtime Prediction with Conformal Matrix Completion

Silverline: Lightweight Virtualization and Orchestration of Distributed Systems
Arjun Ramesh, Tianshu Huang, Emily Ruppel, Dakshina Dasari, Behnaz Pourmohseni, Fedor Smirnov, Marco Giani, Paolo Pazzaglia, Charles Shelton, Nuno Pereira, Arne Hamann, Dirk Ziegenbein, Anthony Rowe
RTAS 2025

We develop Silverline: a framework for programming heterogeneous dynamic distributed systems. Using WebAssembly as a lightweight virtualization framework, we design silverline to be fault-tolerant and real-time, and show it in use on industrial-grade systems.

Silverline: Lightweight Virtualization and Orchestration of Distributed Systems

Empowering WebAssembly with Thin Kernel Interfaces
Arjun Ramesh, Tianshu Huang, Ben Titzer, Anthony Rowe
EuroSys 2025 | Artifact Available — Artifact Functional — Results Reproduced

Webassembly (Wasm) adoption for new domains is often hindered by the lack of standard system interfaces. We propose directly exposing OS userspace syscalls as a compromise.

[Github][Rust Target]

Empowering WebAssembly with Thin Kernel Interfaces

DART: Implicit Doppler Tomography for Radar Novel View Synthesis
Tianshu Huang, John Miller, Akarsh Prabhakara, Tao Jin, Tarana Laroia, Zico Kolter, Anthony Rowe
CVPR 2024, Oral Presentation — Top 0.78% (90) of all submissions

We propose DART — Doppler Aided Radar Tomography, a Neural Radiance Field-inspired approach to radar novel view synthesis using implicit neural inverse imaging.

[Paper][Implementation][Data Collection][Dataset]

DART: Implicit Doppler Tomography for Radar Novel View Synthesis

Optimizer Amalgamation
Tianshu Huang, Tianlong Chen, Sijia Liu, Shiyu Chang, Lisa Amini, Zhangyang Wang
ICLR 2022

While meta-training optimizers from scratch is difficult, much attention has been directed towards developing a diverse body of analytical optimizers. Can we use these optimizers to meta-train a stronger optimizer?

[Poster][Github]

Optimizer Amalgamation

Conference & Invited Talks

Casual Presentations