Me

Taco de Wolff

Data Scientist

Taco de Wolff

As a data scientist, I work at DeWolff Consultancy based in Santiago, Chile, as well as The Netherlands on machine learning research and applications related to climate change.

About me

I am a physicist (and now data scientist) who enjoys working on a wide range of topics and research. Currently I am working as an independent AI/ML consultant on a wide range of data science projects. Before that, I worked two years at the INRIA Chile Research Centre which is part of INRIA, the national research institute in France and a leader in machine learning. Before that, from 2019 till 2021 I worked within the maths department at the Universidad de Chile in Santiago, where our principal work was on Gaussian processes including the development of a Python library. In 2018 I've worked at the bio-engineering department in Auckland, New Zealand, where we've worked with CT-scans of lungs from humans and animals to perform finite element analysis and simulations. Additionally, I do a lot of programming work and I occasionally give tech talks. My interests also extend to exploring macroeconomics and geopolitical ideas, as well as learning new languages and encountering new cultures. I'm an avid kiteboarder riding at spots around the world as much as I can!

Graduated in January 2018 from the Rijksuniversiteit Groningen, I am a master in Physics. My bachelor thesis was about growing graphene on copper foil and for my master degree I specialized in materials research and worked on copper nanoparticle sputtering to create hydrophobic surfaces. Additionally, in Florianópolis, Brazil, I've worked on modelling material deposition using a laser cladding setup with applications in the aerospace industry.

As a programmer, I work as an experienced consultant in programming languages such as Go, PHP, C#, and others, and develop applications for clients. In the meantime I've worked on an assortment of applications and libraries, such as a very fast minifier for web file formats and canvas a 2D vector drawing library with research-backed implementations on flattening and arc length approximations.

Languages and fluency

Publications

Computationally-efficient initialisation of GPs: The generalised variogram method

Universidad de Chile & INRIA Chile Research Centre, Abril 2023

Gaussian process imputation of multiple financial series

Universidad de Chile, February 2020

MOGPTK: The Multi-Output Gaussian Process Toolkit

Universidad de Chile, February 2020

Volume and postural effects on tissue density distribution and tissue mechanics in the porcine lung

University of Auckland, January 2020

Wetting of Copper Nanoparticles covered Silicon/Copper surfaces

Master thesis, Rijksuniversiteit Groningen, May 2017

Graphene CVD growth procedure optimization

Bachelor thesis, Rijksuniversiteit Groningen, July 2011

Projects

Multi-output Gaussian process toolkit

The Multi-Output Gaussian Process Toolkit is a Python toolkit for training and interpreting Gaussian process models with multiple data channels. It builds upon PyTorch to provide an easy way to train multi-output models effectively on CPUs and GPUs. The main authors are Taco de Wolff, Alejandro Cuevas, and Felipe Tobar as part of the Center for Mathematical Modelling at the University of Chile.

Canvas

Canvas is a common vector drawing target that can output SVG, PDF, EPS, raster images (PNG, JPG, GIF, ...), HTML Canvas through WASM, OpenGL, and Gio. It has a wide range of path manipulation functionality such as flattening, stroking and dashing implemented. Additionally, it has a text formatter and embeds and subsets fonts (TTF, OTF, WOFF, WOFF2, or EOT) or converts them to outlines. It can be considered a Cairo or node-canvas alternative in Go.

Minify

Minify is a minifier package written in Go. It provides HTML5, CSS3, JS, JSON, SVG and XML minifiers and an interface to implement any other minifier. Minification is the process of removing bytes from a file (such as whitespace) without changing its output and therefore shrinking its size and speeding up transmission over the internet and possibly parsing. The implemented minifiers are designed for high performance.

Posts

Kiting in Cape Town

May 29, 2019

A review of Cape Town and its kitespots, the big air spot in South Africa.

The perfect webserver

May 28, 2019

Set up a linux webserver using NGINX, PHP-FPM, and Let's Encrypt.

Arc length parametrization

May 25, 2019

Overview of techniques to approach arc length parametrization of Béziers and elliptical arcs.

Kiting in Auckland

November 07, 2018

A review of Auckland and its kitespots, the biggest city in New Zealand.

Kiting in Florianópolis

February 25, 2018

A review of Florianópolis and its kitespots, an island along the coast in southern Brazil.

See the archives for older posts.