London, England, United Kingdom
NewDay is one of the UK’s largest credit card lenders, specialising in the sub-prime market through their own cards as well as prime products in association with brands such as Amazon. I worked in the credit risk data science team, which is responsible for building and maintaining models for fraud, acquisitions, customer management and collections.
MSc in Theoretical Physics with a focus on Quantum Foundations. Master's thesis in Quantum Collapse Models
Worked in the core engineering group which created and maintained the core algorithms underlying the products Researched and implemented algorithms to improve the existing products, working in particular on Spire Talent Acquisition Used machine learning techniques including natural language processing (NLP) and neural networks Worked with scipy, pandas and tensorflow to train a neural network based on word2vec (an NLP algorithm) designed to learn hiring patterns of firms in IT services industry Combined NLP techniques with Collaborative Filtering to train the algorithm to match candidate resumes to job descriptions (i.e. advertised vacancies) The algorithm learned from historic hiring data: searching for keywords in vacancy postings and in resumes that were considered successful and unsuccessful for each vacancy Rolled out a service to consumers which provided context-based synonyms for role names and could map a likely career trajectory from a starting point - This used a graph database called arangoDB with underlying architecture built in java - Database statistics compiled from calls to elasticsearch and word2vec data
Undertook a machine learning project to extract the laws of physics from experimental data and involves genetic programming and genetic algorithms: • Applied a genetic programming technique to predict future dynamics of a physical system with symbolic regression technique • Written in clojure and mathematica, comprises of a genetic algorithm and a polynomial optimiser as the search space was not smooth enough with symbolic regression • Successfully modelled simple harmonic oscillator, coupled harmonic oscillators amongst other physical systems