Phillip Busko
Washington DC
adbv8u@r.postjobfree.com
Github
Hobby Site
Data scientist and software engineer with background in physics, data mining, database administration, and web development. Interested in all realms of artificial intelligence and modelling; including machine learning, scientific analysis, game AI, computer vision, neural networks, NLP, robotics, etc.
Technical Skills
Python, SQL, C#, Javascript, HTML, SASS
Pandas, Scikit-learn, Plotly, XGBoost, Django, React
Visual Studio Code, Jupyter, MS SQL Server, PostgreSQL, AWS Console
Technical Projects
Filmophile - Github
Scraped data from movie information and streaming service websites with selenium package
Engineered several features as a mix of the same column from different data sources, applying one-hot and multi-hot encoding to categorical columns
Searched for the best classifier among: logistic regression, naïve bayes, KNN, random forest, SVM, xgboost
Tuned support vector machine (sklearn sgdclassifier) to improve accuracy by 6%
Regression for Pricing of Lego Sets - Github
Scraped Lego set data with requests and beautifulsoup libraries
Tuned a linear regression regularization model to predict the price of a new Lego set with a R2 of 0.86
Preserved the interpretability of beta-coefficients in order to construct a symbolic mathematical model
Classification for Profitability of New Lego Sets - Github
Processed the columns of the Lego data set to engineer new coarse-grain features and the target variable
Trained logistic regression with different solvers and regularization for a baseline accuracy of 78%
Tuned a random forest with grid search of hyper-parameters to improve accuracy to 80%
Changed the balance of input classes with the imblearn pipeline using under sampling and SMOTE
Professional Experience
Data Science Fellow, Flatiron School, Washington DC
2019 – 2019
Learned fundamentals of data science practice and theory using Python in Jupyter notebooks
Studied pandas dataframe, website scraping and APIs, exploratory data analysis, statistical inference, supervised learning classifiers, unsupervised learning clustering, and natural language processing
Collaborated on 7 different projects, each focusing on a new area of study
Software Engineer, IT Innovative Solutions, Gaithersburg MD
2018 - 2019
Used SQL Server Management Studio to import data files and run processing scripts
Learned React, Angular and SASS to build client-side applications
Implemented a RESTful server’s API endpoints in C#
Founder & CTO, Lige Ma, Baltimore MD
2015 - 2017
Self-taught full stack development and deployment: Django, PostgreSQL, jQuery, Amazon Web Services
Website features include web crawlers for soccer data, user authentication, global chat using web sockets, PayPal monetization scheme, user-to-user messaging, and thin client paradigm for the game state
Software Engineer, Audacious Inquiry, Baltimore MD
2012 - 2014
Learned SQL and SSMS and ultimately created stored procedures dynamically
Increased the correctness of synthesized HL7 inpatient hospital visits from 60% to 90%
Administered databases (2NF, star schema) utilizing diagrams, partitioning and backups
Built extract-transform-load (ETL) packages for loading databases and processing existing data
Authored extensive software documentation to on-board new personnel
Engineer, Sensor Concepts & Applications, Glen Arm MD
2010 - 2012
Learned C# and Visual Studio, and began contributing to production code within 3 weeks
Parsed various vendor data file formats, including xml and binary, and converted them to a common data structure for visualization and analysis of radiation spectra
Created MCNP (Department of Energy tool) high-energy physics simulations that mirrored real-life experiments
Education
Boston University, Boston, MA
Masters Physics – May 2008
University of Maryland, College Park, MD
B.S. Physics, magna cum laude – May 2004
B.S. Computer Science – May 2004