• Simulated Environments

  • Deep Learning

  • Data Analytics

  • High Performance Computing

About Me

Graduating with a First Class Honours in Computer Science and subsequently being awarded a full PhD scholarship for research in Deep Learning and Big Data analytics. An outgoing researcher who wishes to continue developing technical skills and new approaches while providing meaningful real-world applications. Extensive range of relevant technical ability developed throughout time in academia. Highly motivated and hardworking, consistently looking to better oneself and others. Driven by self-learning and the gripping desire for completion.

Throughout my research as a PhD student, I have taken a highly experimental stance, pushing for automated systems that can simultaneously output results while I further my knowledge in topic areas. In post graduate and PhD studies multiple peer reviewed publications in many different fields. A vast array of programming language applications and corresponding tools, from python to c++ using Github, SVN and more for version control and safe collaborations.



Quick development and prototyping experiments. Extensive use for deep learning in particular applications of convolutional and multilayered neural networks, graphing and statistical analysis.


Uses for highly optimized and niche programming solutions; such as high-performance computing and graphical simulations.


Large time spent using C# at undergraduate, large modular programs utilising abstraction and polymorphic behaviour.


Previous uses include androird mobile app development and game automation tools.


Used in both website development and light-weight web driven game creation.


Deep Learning

The main focus of my PhD work as present, with a direct interest in image processing through neural networks, both discrete and continuous outputs. Future work aims analyse procedural content and synthetic images using generative models.

High Performance Computing

Research internship generating frameworks to process data through The University of Hull's supercomputer, Viper. Large-scale processing of matrix operations using Cuda and multiple high-end GPUs for deep learning.


Dissertation project for undergraduate creating a simulation with the ability to mimic real-world terrains. Providing a huge scale procedural planet which can be traversed in real time.

Data Mining and Analytics

Throughout projects involving large-scale data, consistent access and use of data analysis tools to justify, graph and prepare data sets for future work. Using various supervised and unsupervised techniques to find hidden representations in data sets. Readable visualizations of data through heat maps and 3D plots.


List of academic posters and paper publications; both first and co-authored

Deep Learning for Wave Height Classification in Satellite Images for Offshore Wind Access

Summarising preliminary experiments and poster publications into a broader collection of writing on the methodologies involved with; processing large amounts of open source satellite data, cleaning and preparation of the data set and the continuous classification of a wave height for a given satellite image.

Utilising VIPER for Parameter Space Exploration in Agent-Based Wealth Distribution Models

Developing a supercomputing framework with the ability to exhaust parameters in a model or simulation effectively.


Toolkit Development for Parallelized Agent Based Wealth Distribution Models

Practical application of previous parameter exploration framework tools, simulation of an agent-based model aiming to reenact how financial population segregation occurs.


Procedural Generation using Spatial GANs for Region-Specific Learning of Elevation Data

A robust method for the learning and generation of elevation data in the form of a height map, using a type of GAN that learns non-spatially dependant features.


Time to Die: Death Prediction in Dota 2 using Deep Learning

Using a multilayered perceptron trained on a vast array of features, that utilizes shared weights within the first half of the network has the ability to train a model that can accurately predict the death-state of every player within a live or recorded game of Dota 2.


Naive Mesh-to-Mesh Coloured Model Generation using 3D GANs

Learning of pre-exsisting mesh data, using voxel downsampling to train a 3D GAN with 4 channels to include both colour and position. Marching cubes is then used to convert the lkearned voxel back to mesh format.


Realistic and Textured Terrain Generation using GANs

Generation of terrain maps from pre-exsisting satilitte data, with the inclusion of texture data, trained on a non-spatially dependant GAN .


Human Point Cloud Generation using Deep Learning

Collaboration with Sony, exploring dense correspondence point data for controlled generation of human pose and animation.


Illuminating Game Space Using MAP-Elites for Assisting Video Game Design



Please email me with any enquiries ryan.spick@hotmail.co.uk