Samuel Kelly

Data Scientist - Professional Development @GitHub

London, GB

Professional Development: upskilling in Python, maths, and machine learning.· Developed proficiency in Python, including the NumPy, Pandas, SciPy, Matplotlib and Scikit-learn libraries.· Gained fluency in SQL and Git.· Re-learnt Maths A-Level and studied linear algebra for data science.Created 2 projects exemplifying my skills.US Flight Delay Project:I predicted the delay of commercial flights using large-scale categorical data. · Standardised heterogeneous airport identifiers using my data engineering skills and engaged in domaindriven data cleaning to improve the data integrity of time-related columns. · Predicted flight delays by optimising linear regression and gradient boosting models, leveraging their abilities to efficiently handle large amounts of categorical data. · Found insights through feature importance analysis. · Analysed the residuals and biases for several ML models to quantify their effectiveness. · Used Seaborn and Matplotlib for data visualisation in my EDA to uncover variations and explain key patterns in the data with real world implications.Esports Predictions Project:I built a model to predict the results of professional Esports games using historical data, a custom Elo system, and machine learning algorithms.· Developed a custom Elo system incorporating 6 component ratings to model each team’s relative performance over time. · Optimised the Elo system using SciPy, and used NumPy’s matrix operations to improve efficiency. · Engineered model-based dynamic features from Elo ratings to improve the predictive performance of downstream models. · Utilised meta-modelling to prevent underfitting and to more accurately predict game results. · Analysed model calibration using Brier scores, ECE and bootstrapping.

The University of Manchester

Bachelor of Arts - BA, Philosophy and Religious Studies