Matthew Trentacoste

I'm a research generalist, equally at home working as engineer, researcher, and data scientist. I work end-to-end transforming research ideas into usable technology. Broad experience in domains including graphics & imaging, machine learning, data analysis, GPU programming, web backend, and mobile development.

TL;DR

imaging

PhD in image processing and computer graphics. Implemented algorithms and pipelines, including early HDR imaging work. Shipped a top 10 iOS photography app.

ml

Deployed demand-forecasting models into production. In addition to training the model, built pipeline & hosting infrastructure. Maintained dispatch systems that relied on the models.

GPUs

Implemented GPU numerical linear algebra libraries and tiled memory managers in pre-CUDA shaders. Implemented graph-based image processing framework capable of decomposing imaging operations over multiple passes.

cameras

Calibrated image sensors, both color and luminance. Experience with machine vision camera SDKs.

eng

10+ years experience with C++, Python. OpenGL, OpenCV.

Professional experience

2018-now

Research Scientist, Facebook, Menlo Park, CA.

Various projects related to computer vision in the Applied AI group.

2017

Consulting Data Scientist, Figma, San Francisco, CA.

Establishing the analytics foundation. Logging, infrastructure, and analysis assisting both growth and product decisions.

2014-2016

Data Scientist, Sprig, San Francisco, CA.

Involved in all data aspects of a complex logistics and culinary business. Designed and implemented real-time system for accepting orders, intelligently throttling order-rate based on system load to maximize delivery efficiency while preserving SLA. Designed demand forecasting tools for planning and procurement of future menus. Provided deep analysis to product and business teams.

2013-2014

Computer Scientist, Adobe, San Francisco, CA.

Designed and implemented mobile photography apps, covering GPU-processing frameworks, implementing image processing algorithms, and desiging interfaces to expose those algorithms. Developing data pipelines for creative products. Conducting research identifying and classifying visual style of images, as well as recommending content-specific edits.

2011-2012

Researcher / Developer, Pocket Pixels, Vancouver, BC.

Top-selling selective desaturation photography for iPhone and iPad. Completely rebuilt the app’s graphics framework, included interactive image adjustments and faster performance. After release, the app rose to #6 in the US store.

2009-2011

Scientist, Empty Research, Vancouver, BC.

Multiple projects including computer vision and artificial intelligence algorithms for computational aesthetics, automatically generated websites for mobile devices, and GPU-accelerated mobile photography and vision applications.

2007-2009

Research Engineer, Dolby, Vancouver, BC.

Supported the development of Dolby technology licensed to manufacturers, including the SIM2 Grand Cinema SOLAR. Developed calibration methods for accurate display of content on HDR displays, and environment sensing setups for adjusting display color to the viewing environment. Reduced visible artifacts from LED variation by 90%.

2004-2007

Principal Software Developer / Researcher, BrightSide Technologies, Vancouver, BC.

Responsible for the complete design, development, and implementation of a the video processing pipeline for LED-backlight display prototype, the BrightSide DR37-P. Developed the GPU-accelerated software used for all demos, securing 2 years worth of investment capital. Created third-party API for display hardware. Expanded IP portfolio.

Education

2012

PhD in Computer Science, University of British Columbia

Manipulating Scale-Dependent Perception of Images, advised by Wolfgang Heidrich.

2006

MSc in Computer Science, University of British Columbia

Photometric Image Processing for High Dynamic Range Displays, advised by Wolfgang Heidrich.

2003

BSc in Computer Science, Carnegie Mellon University

Implementing Performance Numerical Libraries on Graphics Hardware, advised by Doug James.

Journal articles & papers

2014

Recognizing Image Style, S. Karayev, M. Trentacoste, H. Han, A. Agarwala, T. Darrell, A. Hertzmann, H. Winnemoeller. British Machine Vision Conf.

2012

Scale-Dependent Perception of Countershading: Enhancement or Artifact?, M. Trentacoste, R. Mantiuk, W. Heidrich, F. Dufrot. Eurographics

2011

Glare Encoding of High Dynamic Range Images, M. Rouf, R. Mantiuk, W. Heidrich, M. Trentacoste, C. Lau. CVPR

Blur-Aware Image Downsizing, M. Trentacoste, R. Mantiuk, W. Heidrich. Eurographics

2010

Defocus Techniques for Camera Dynamic Range Expansion, M. Trentacoste, C. Lau, M. Rouf, R. Mantiuk, W. Heidrich. Proceedings of Human Vision and Electronic Imaging XXI

2007

Photometric Image Processing for High Dynamic Range Displays, M. Trentacoste, W. Heidrich, L. Whitehead, H. Seetzen, G. Ward. Journal of Visual Communication and Image Representation, Special Issue on High Dynamic Range Imaging

Ldr2Hdr: On-the-fly Reverse Tone Mapping of Legacy Video and Photographs, A. G. Rempel, M. Trentacoste, H. Seetzen, D. Young, W. Heidrich, L. Whitehead, G. Ward. Transactions on Graphics, SIGGRAPH

2005

Real Illumination from Virtual Environments, A. Ghosh, M. Trentacoste, H. Seetzen, and W. Heidrich. Eurographics Symposium on Rendering

Volume Rendering for High Dynamic Range Displays, A. Ghosh, M. Trentacoste and W. Heidrich. International Workshop on Volume Graphics

2004

High Dynamic Range Display Systems, H. Seetzen, W. Heidrich, W. Stuerzlinger, G. Ward, L. Whitehead, M. Trentacoste, A. Ghosh, A. Vorozcovs. Transactions on Graphics, SIGGRAPH

Patents

2009

Mitigation of LCD Flare, G. Ward, J. Harrison, H. Seetzen, M. Trentacoste, US 2010/0277515 A1

2007

Multiple Modulator Displays and Related Methods, W. Heidrich, M. Trentacoste, G. Ward, H. Seetzen, US 2010/0091045 A1

Talks & miscellany

2015

Data Science Community Founding Member, Greylock Partners

Adobe Photoshop Fix

2013

Image Features in Python talk (slides) (video)

2012

Color Splash

2011

Manipulating Scale-Dependent Perception of Images invited talk. Disney Research, Massachusetts Institute of Technology, NVIDIA

Blur-Aware Image Downsizing invited talk. Bangor University, Adobe

2009

Dolby Vision display

2005

High Dynamic Range Techniques in Graphics: from Acquisition to Display. M. Goesele, W. Heidrich, B. Hoefflinger, G. Krawczyk, K. Myszkowski, M. Trentacoste, Eurographics

Technical skills

experience

image processing & sensors, graphics & rendering, data analysis, machine learning, computer vision, data visualization, GPU programming, visual perception, iOS development

languages

C/C++, Python, Objective-C, Ruby

graphics

OpenGL, OpenCV, GLSL shaders, CUDA

data

NumPy, SciPy, PyData ecosystem, SQL

ml

scikit-learn, Keras

web

AWS, Heroku, Flask, Javascript, HTML, CSS

References

Ask and ye shall receive.