Website | Github | [email protected]
Summary
Software engineer specialised in machine learning with six years experience in the software industry for London-based startups. Strong software engineering professional skilled in PyTorch, Keras, Tensorflow, cloud-computing, prototyping, and Amazon Web Services (AWS). I have worked on acclaimed applications including two Apple Apps of the Year, a Webby, and a BAFTA winner.
Software Projects
Personal Website: www.alanmartyn.com (for documentation, source code and posts)
Deep Learning Paper Implementations
- Implemented the Pix2Pix GAN generative model using Keras and Tensorflow, then trained the model to produce photos of architectural facades from diagrams.
- Implemented the AWD-LSTM NLP language model using PyTorch, wrote the LSTM cell from scratch, and fixed GPU memory-leaks.
- Implemented three U-net architectures for image segmentation using Keras and Tensorflow, then compared performance on the ISBI Challenge dataset.
- Implemented the ResNet convolutional architecture in PyTorch with configurable depth, then reproduced the paper's CIFAR-10 vision experiment.
Machine Learning Projects
- Dataset Design: Used web-scraping and parallel processing to build a 20,000 image dataset. Trained generative models for textile designs.
- Vision: Entered the Human Protein Atlas Image Classification Kaggle competition. Ranked in the top 23% of 2,200 entrants in collaboration with a team-mate.
- Prototyping: Competed in the South East Water, Preventing Bursts Hackathon with a team-mate. Pre-processed a large commercial dataset, and trained a tree-based model. Ranked top-4 overall from 15+ commercial entrants.
- Feature Engineering: Applied feature engineering with a simple linear-regression model to the House Prices: Advanced Regression Techniques Kaggle challenge.