https://github.com/atsai24/song_recommenderUsed the Million Song Dataset to build a song recommender web application where user inputs their favorite songs and are recommended 5 new songs based off of those inputsTransformed data set of user-song interactions into a large sparse matrix using the SciPy and Pandas packagesUsed LightFM package to build an implicit recommendation model based on user-song interactions as well as song tags as item featuresLaunched AWS EC2 instance and S3 bucket to efficiently perform computationally heavy tasks such as large data transformations and model fitting.Created a local web application using Flask that runs user inputs through model to output 5 recommended songs.

https://github.com/atsai24/music_audio_analysisAnalyzed a database of over 100,000 tracks and built a model to classify their genre, using metadata, audio features, and Spotify’s song features. Analyzed track audio and visualized it using waveform and spectrograms using the Librosa and Matplotlib Python packages.Examined the differences between the 12 genres by grouping and visualizing audio features for each genre.Tested various models including neural network, gradient boosting, and random forest, and decided on the most efficient random forest model that predicted genre with 75% accuracy.Created docker container to access Keras/Tensorflow package for neural network testing.

https://github.com/atsai24/Time-for-TennisExplored the amount of time professional tennis players took to serve, and quantified the psychological factors of the sport by examining the differences in serve times when the server was under pressureAnalyzed and visualized dataset of serve times using the Pandas, Numpy, Matplotlib, and Seaborn Python packages, and split data by high-pressure (game-point) and not high pressure (not game-point).Ran a statistical analysis and hypothesis test on the serve times using the Scikit-Learn Python package, and found a statistically significant increase in time for high-pressure points.

Improved company workflow by building and customizing a CRM system using Bitrix24Developed an understanding of the company’s business process by interacting in Mandarin with employees from Sales, Marketing, and HR, in order to best integrate the CRM system to their needsAutomated various business processes such as creating leads, assigning sales employees to those leads, sending invoices and quotes, and tracking revenue flowCleaned database of customers by removing duplicate customers, removing false leads, and updating contact information