Stuart Davie

AI • Science • Philosophy • Data • Business

About Me

I'm a scientist with passion for research and development. Currently my focus is in two areas - Data Science and AI - although I keep a keen interest in various other fields of mathematics, physics, computer science, and philosophy. Day to day, I am busy at Peak developing our Artificial Intelligence System, designed to help other companies do great things with data. I have created this website to serve as a place where I can share a little bit about the projects I am working on; keep a blog regards AI and Data Science; get some web development practice; and where people can get in touch with me.

After a dual major in Physics and Mathematics at Griffith University, I completed a post-graduate research year calculating advection and diffusion in the human respiratory system using differential calculus and the Finite Volume Method. I then completed a PhD in statistical mechanics/physical chemistry in the Bernhardt group at the Queensland Micro- an Nanotechnology Centre, calculating free energy differences in non-equilibrium systems, using the work relations. After this, I moved to Manchester, UK, with my wife to join the Popelier research group, developing the worlds first entirely machine learned force-field for chemical simulation. Whilst I still enjoy collaborating with past groups, I am now working at Peak, as a data scientist, responsible for research and developmenta, where I can apply the diverse skills I have learned from the natural and formal sciences to the transformation of data big and small into business insight and solutons.

To connect on LinkedIn, GitHub, or Google Scholar, check out the links below.

linkedin github

Projects & Research

Selected Projects

MPI_PSO

A Fortran MPI implementation of Particle Swarm Optimization. This program was written to serve as a working example to assist Daresbury Labs

FFLUX Tools

Some of the tools developed during the development of FFLUX - the first fully machine learned force field. FFLUX is a large project with many moving pieces, but I intend to slowly build this up with relevant code.

R Highcharter Flexadash

An example dashboard, whipped up in R, using Highcharter and Flexadash

Machine learning atomistic properties

The accuracy of ab initio calculations without ab initio calculations for charged systems: Kriging predictions of atomistic properties for ions in aqueous solutions
The application of machine learning to the prediction of ionic, atomistic properties.

Molecular Machine Learning

Prediction of intramolecular polarization of aromatic amino acids using kriging machine learning
Application of Gaussian Process Regression (Kriging) to the intra-molecular, inter-atomic electrostatic energies of all natural aromatic amino acids.

High Performance Machine Learning

FEREBUS: Highly parallelized engine for kriging training
Application of OpenMP and MPI parallelisation paradigms to hyperparamter optimisation for Gaussian Process Regression machine learning algorithm. Particle Swarm Optimisation and Differential Evolution are used as optimisation algorithms

Molecular Feature Engineering

Incorporation of local structure into kriging models for the prediction of atomistic properties in the water decamer
Incorporation of local structure into kriging models for the prediction of atomistic properties in the water decamer

Stochastic Optimisation Algorithms

Optimization algorithms in optimal predictions of atomistic properties by kriging
The application of Particle Swarm Optimisation and Differential Evolution (a type of Genetic Algorithm) to the optimisation of Machine Learning (Gaussian Process Regression) hyper-parameters.

Use of a machine-interpolated potential for molecular geometry optimisation

Geometry Optimization with Machine Trained Topological Atoms
Distorted water monomers are successfully optimised using a fully machine-learned forcefield.

Handling a variable number of features for machine learning in molecular systems

Kriging atomic properties with a variable number of inputs
A novel methodology is presented, which removes the limitations of requiring different models for different sized systems.

Long-range Water Interaction Energies

The long-range convergence of the energetic properties of the water monomer in bulk water at room temperature
Investigation into the long range properties of a water monomer, in a box of water. Largest Quantum Theory of Atoms in Molecules partitioning to date.

Free Energy calculations in systems changing density

The free energy of expansion and contraction: treatment of arbitrary systems using the Jarzynski equality
The application of the Work Relations, namely the Jarzynski Equality and a Maximum Likelihood Estimator, to free energy predictions in systems expanding and contracting.

Increasing free energy calculation efficiency through non-equilibrium transformations

Free Energy Calculations with Reduced Potential Cutoff Radii
The application of the free energy work relations to systems undergoing changes to their potential, greatly computational increasing efficiency.

Free energy calculations from stochastic and deterministic simulations

Applicability of optimal protocols and the Jarzynski equality
Comparison of stochastic and deterministic transformations for free energy calculations.

Relative Free Energies from Non-Equilibrium Simulations

Thesis: Relative Free Energies from Non-Equilibrium Simulations: Application to Changes in Density

How to use Machine Learning for Customer Acquisition

Supervised customer segmentation, conversion pathways, and recommendation systems

Using machine learning to monitor company risks

How a combination of web-scraping and Natural Language Processing can be combined to create a real time system for monitoring comapny risk

Free Energy Calculations in Phase Transition Systems

Free energy calculations in systems changing phase, with changing potentials

Free Energy Calculations in Phase Change Systems

Application of non-equilibrium work relations to systems under compression and expansion


Current Research Interests

  • Optimal imputation methods for multi-domain data sets
  • Automatic data relation detection
  • The application of Named Entity Recognition and domain partitioning for NLP driven web monitoring
  • Automatic corpus generation for PPC applications
  • ANN-applied image feature-extraction for marketing optimisation
  • Consumer lifetime modelling
  • Business lead optimisation
  • A Classical Exorcism of Maxwell’s Demon from Liboff’s Tri-Channel
  • Improved Subset Selection from Large Datasets for Gaussian Process Modelling
  • Application of Gaussian Processes to Micro-solvated Systems
  • Microcanonical Simulation of Small Molecules using Machine Learned Force-Fields
  • Optimization of Small Molecules using Machine Learned Force-Fields

Contact Me



Back to Top