Shing Chan, Developer in Asuncion, Paraguay
Shing is available for hire
Hire Shing

Shing Chan

Verified Expert  in Engineering

Machine Learning Developer

Asuncion, Paraguay

Toptal member since October 27, 2021

Bio

Shing is a researcher and developer of machine learning methods with experience across diverse domains, including physics-informed machine learning for fluid simulations, generative AI for geomodelling, clinical risk scores for patient screening, customer lifetime value models for eCommerce, and models for sports betting and algorithmic trading. Shing currently works at the University of Oxford, researching methods to leverage wearable sensor data for health analytics.

Portfolio

University of Oxford
Machine Learning, Deep Learning, Time Series Analysis, Wearables, Bash...
KEG Systems LLC
Machine Learning, Data Science, Trading, TensorFlow, PyTorch, Keras, Git, Linux...
Heriot-Watt University
Computational Fluid Dynamics (CFD), Machine Learning, Computer Vision...

Experience

Availability

Part-time

Preferred Environment

Linux, Git, PyTorch, Keras, Bash, Fortran, Vim Text Editor, Amazon Web Services (AWS), Python

The most amazing...

...thing I've developed is a generative AI for geomodelling, published in CompGeosci 2019.

Work Experience

Researcher

2019 - PRESENT
University of Oxford
  • Developed algorithms for wearable devices to recognize sleep, walking, cycling, and other activities.
  • Applied time-to-event analysis (e.g., Cox PH, survival forests) to model time to hospitalization or death based on patient characteristics and alternative data such as wearable sensor data.
  • Developed a deep learning model for data-driven information extraction from high-res and unstructured wearable sensor data for health and behavioral insights.
  • Created Python packages to process and analyze wearable sensor data for health analytics.
  • Built data pipelines to perform time-series analyses on terabytes of healthcare data.
Technologies: Machine Learning, Deep Learning, Time Series Analysis, Wearables, Bash, Fitness Trackers, Risk Models, Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Signal Processing, Python, TensorFlow, Keras, Git, Linux, NLP, Scientific Software, Health, Time Series, R, Data Science, Predictive Modeling, SQL, PostgreSQL, Data Mining, Artificial Intelligence, Algorithms

Machine Learning Expert

2018 - 2019
KEG Systems LLC
  • Researched novel features (e.g., player-player, player-team, team-team interaction features) to predict game outcomes for sports betting (e.g., money line, over-under, and spread), emphasizing calibration to inform bet sizing and risk management.
  • Created reproducible pipelines for daily retraining, including feature selection, fine-tuning, and pruning.
  • Oversaw deployment and decision-making, betting with real money and tweaking metamodels based on feedback.
Technologies: Machine Learning, Data Science, Trading, TensorFlow, PyTorch, Keras, Git, Linux, Signal Processing, Python, NLP, Sports, Scientific Software, PostgreSQL, Time Series, Algorithms, Gambling, Data Science, Predictive Modeling, SQL, Data Mining, Artificial Intelligence, Algorithms

PhD Candidate

2015 - 2018
Heriot-Watt University
  • Developed a physics-informed machine learning model to speed up computationally expensive Monte Carlo fluid simulations.
  • Developed a novel framework for geological reconstruction based on generative models (e.g., GANs, VAEs) to enhance geological realism for improved accuracy of oil production forecasts in Bayesian history matching.
  • Created Python packages for subsurface fluid simulations.
Technologies: Computational Fluid Dynamics (CFD), Machine Learning, Computer Vision, Generative Adversarial Networks (GANs), Variational Autoencoders, Physics Simulations, TensorFlow, PyTorch, Git, Linux, Python, NLP, Scientific Software, Time Series, Data Science, Predictive Modeling, Artificial Intelligence, Algorithms

Engineering Intern

2014 - 2014
FAdeA
  • Assisted in the maintenance and repair of aircraft components.
  • Assessed the capabilities of aircraft repair stations, making sure tools and procedures were in order according to technical manuals.
  • Issued reports documenting deviations from technical manuals, including changes in procedures, the use of original equipment manufacturer (OEM), or refurbished parts.
Technologies: Aerospace & Defense, Aircraft & Airlines, Engineering, Scientific Software, Time Series, Data Science, Predictive Modeling, Artificial Intelligence

Research and Development Intern

2013 - 2014
Instituto Universitario Aeronáutico
  • Contributed to an in-house software for viscous flow simulation, extending it with the arbitrary Lagrangian-Eulerian formulation on unstructured grids.
  • Identified bottlenecks in the simulation software and parallelized them with OpenMP where possible.
  • Ported code sections with CUDA Fortran to enable GPU acceleration, resulting in more than 10 times the speed-up.
Technologies: Computational Fluid Dynamics (CFD), Aerodynamics, Numerical Simulations, Fortran, GPU Computing, CUDA, OpenMP, Physics Simulations, Git, Linux, Python, MATLAB, Scientific Software, Time Series, Data Science, Predictive Modeling, Artificial Intelligence, Algorithms

Package for Processing and Analysis of Wearables' Data for Health Analytics

https://github.com/activityMonitoring/biobankAccelerometerAnalysis
This Python package extracts a wide range of clinically relevant statistics related to activity and sleep patterns obtained from wearable activity trackers. I contributed to all aspects of the software, from signal processing, feature engineering, machine learning to maintenance and packaging.

Numerical Optimization with Natural Evolution Strategies

https://github.com/chanshing/xnes
This is a simple-to-use Python module for optimization via natural evolution strategies. It is ideal for hard optimization problems involving highly non-linear, non-convex objective functions or when gradients are unavailable or difficult to compute.

Synthesis of Geological Images

https://github.com/chanshing/geocondition
This tool is used to generate geological images in a conditional manner from a generative neural network. It is useful for subsurface reconstruction to obtain more realistic geomodels, thus optimizing oil and gas exploration and extraction pipelines.

Physics-informed Machine Learning for Accelerated Simulations

https://www.sciencedirect.com/science/article/abs/pii/S0021999117307933?utm_medium=email
Developed a hybrid method that combines machine learning and physics to speed up fluid simulations. I trained a convolutional neural network to learn fluid dynamics from ground-truth snapshots, in which conservation laws further correct errors. The framework works under a multiscale finite volume formulation.
2015 - 2018

PhD in Petroleum Engineering

Heriot-Watt University - Edinburgh, United Kingdom

2007 - 2014

Engineer's Degree in Aerospace Engineering

Instituto Universitario Aeronáutico - Cordoba, Argentina

SEPTEMBER 2016 - PRESENT

Financial Markets

Yale University | via Coursera

APRIL 2014 - PRESENT

Heterogeneous Parallel Programming

University of Illinois | via Coursera

APRIL 2014 - PRESENT

Programming Mobile Applications for Android Handheld Systems

University of Maryland | via Coursera

Libraries/APIs

PyTorch, Keras, Scikit-learn, TensorFlow, OpenMP

Tools

Git, MATLAB, Vim Text Editor

Languages

Python, Java, Fortran, Bash, C, SQL, R

Platforms

Linux, CUDA, Android, AWS

Paradigms

Parallel Programming

Storage

PostgreSQL

Other

Physics Simulations, Machine Learning, Deep Learning, Time Series Analysis, Generative Adversarial Networks (GANs), Artificial Intelligence, Time Series, Predictive Modeling, Data Science, Algorithms, Algorithms, Signal Processing, Data Mining, NLP, Numerical Methods, Numerical Analysis, Physics, GPU Computing, Computational Fluid Dynamics (CFD), Computer Vision, Data Science, Scientific Software, Finance, Aerodynamics, Numerical Simulations, Optimization, Convolutional Neural Networks (CNNs), Wearables, Fitness Trackers, Risk Models, Variational Autoencoders, Recurrent Neural Networks (RNNs), Health, Aerospace & Defense, Aircraft & Airlines, Engineering, Trading, Gambling, Sports

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring