Watch where you're pointing that thing
Henrique Teles Maia

Ph.D. Candidate at Columbia University

Hello there!


I am a final-year Ph.D. Candidate in Computer Science at Columbia University, advised by Prof. Eitan Grinspun and Prof. Changxi Zheng. My research interests include Physics-based Simulation, Machine Learning, and Computational Fabrication.

I am on the job market this year. If you are interested, please contact me!

Currently, I am interested in merging simulation techniques with machine learning, hoping to trade-off bottlenecks and inaccuracies with data-driven techniques. Machine-learning and model-reduction methods benefit from large demonstrative datasets; the more data present, the easier it becomes to converge on a solution. Aggregating and formatting data is not easy however, examples are not always readily available and the collected samples must be labeled or tagged appropriately. Fortunately, accurate simulations allow for the generation of limitless data under controlled scenarios. With every variable accounted for, the overhead of gathering and preparing the data dissapears. Coming full circle, this data can then be used to train stable and efficient alternatives to physics simulations that can improve time-bound applications or even produce differentiable simulations that can be tuned to achieve specific dynamics.

In the past, I have worked on tagging physical hyperlinks onto intricate everyday objects, as well as exploring side-channel exploits of neural architechtures running on modern GPUs. I also love to tinker, build, and hack. I've been known to augment 3D printers and design award winning robots.

Outside of research you can find me playing ⚽ or 🏐. I also love to cook and explore the food scene around me, feel free to reach out to me for recommendations!

I am originally from São Paulo, Brasil but have spent the last few years living in (and photographing) New York City.

For more information about me, see my resume or contact me.



Education

2018-now
Columbia University

Ph.D. in Computer Science

2021
Columbia University

M.Phil. in Computer Science

2017
Columbia University

M.Sc. in Computer Science

2010-2015
Columbia College and Columbia Engineering

Dual Degree Program, B.A. in Computer Science and B.S. in Mechanical Engineering

Professional Experience

2017
University of Tokyo

Visiting Scholar

Mentored by Yonghao Yue

2017
Disney Animation Studios

Technology and Innovation Research Intern

Mentored by Rasmus Tamstorf

2015-2016
University of Texas, Austin

Visiting Scholar

Mentored by Etienne Vouga

2015
Adobe Systems Inc.

Creative Technologies Lab Intern

Mentored by Danny Kaufman.

2013
1stDibs

Back-end Software Developer




For more details, please see my full CV (PDF).

News



07/2022
Our paper on Mobile Manipulation Leveraging Multiple Views is named finalist for Best Paper Award at IROS 2022!

04/2022
ACM Communications featured our side-channel findings on Neural snooping!

11/2021
Paper accepted (preprint) to USENIX Security conference!

01/2021
Passed my Ph.D. thesis proposal!

08/2019
Presented a talk at SIGGRAPH on tagging arbitrary 3D printed shapes

07/2019
Passed my candidacy exploring Can We Learn to Sim?


Current Projects



Simulations at this scale are visually impressive but painfully slow to produce.

Data Driven Hair Simulation

Advised by Eitan Grinspun and Changxi Zheng


Next-best-view planning improves grasping success in unknown scenarios.

Mobile Manipulation Leveraging Multiple Views

Work with David Watkins-Valls at the Columbia Robotic's Lab

Advised by Peter Allen and Nicholas Waytowich


Networks allow for significant reduction of variables

Continuous Reduced-Order Modeling of PDEs

Work with Peter Chen and G Pershing


A simple baseline sim for experimentation.

GPU Discrete Elastic Rod Library

Work with Peter Chen and Mingxuan Li

Advised by Eitan Grinspun


Papers



CROM: Continuous Reduced-Order Modeling of PDEs Using Implicit Neural Representations

Peter Yichen Chen, Jinxu Xiang, Dong Heon Cho, G A Pershing, Henrique Teles Maia, Maurizio Chiaramonte, Kevin Carlberg, Eitan Grinspun
Submitted to NeurIPS 2022
[ArXiv]

Mobile Manipulation Leveraging Multiple Views

David Watkins-Valls, Henrique Maia, Jacob Varley, Madhavan Seshadri, Jonathan Sanabria, Nicholas Waytowich, Peter Allen
Accepted to IROS 2022, Best Paper Award Finalist

Can one hear the shape of a neural network?:

Snooping the GPU via Magnetic Side Channel

Henrique Teles Maia, Chang Xiao, Dingzeyu Li, Eitan Grinspun, Changxi Zheng
Accepted to USENIX Security (2022)

LayerCode: Optical Barcodes for 3D Printed Shapes

Henrique Maia, Dingzeyu Li, Yuan Yang, Changxi Zheng
ACM SIGGRAPH 2019

A Multi-Scale Model for Simulating Liquid-Hair Interactions

Yun (Raymond) Fei, Henrique Maia, Christopher Batty, Changxi Zheng, Eitan Grinspun
ACM SIGGRAPH 2017

Bedrock erosion by sliding wear in channelized granular flow

C. Hung, C.P.Stark, H. Capart, B. Smith, H. Teles Maia, L Li and M. Reitz
American Geophysical Union 2014
[Paper]

Landslide dynamics from seismology and simulation

C. P. Stark, C. Hibert, G. Ekstrom, M. Reitz, B. Smith, E. Grinspun, H. Teles Maia, and D. Kaufman
Modeling Granular Media Across Scales 2014

Contact


Henrique Teles Maia
E-mail: <x>@cs.columbia.edu, where x=henrique.
Office: CEPSR 602 (map).

Extra: My first name (Henrique) is pronounced as n-he-key ( /ẽ.ˈʁi.ki/ ).
You can also try this: