Client: The Globe and Mail – Recommender System• Researched and tested various content-based filtering methods to find an optimal base for the recommender system • Participated in developing a pipeline in Databricks using PySpark to retrieve clickstream data and articles for recommender system• Developed multiple content-based recommender systems on Databricks using BERT, TF-IDF, and DAN and collaborated with the editorial team to identify the final system using A/B testing• Automated the recommender system on Databricks to create recommendations daily and save to a DynamoDB database• Presented updates and project progress to Data Science team manager on a weekly basis, as well as a formal presentation to the larger globe and mail team upon the completion of the recommender systemClient: The Globe and Mail – Storyline Summarizer• Researched and tested extractive and abstractive summarizer models to find base for Storyline Summarizer• Developed an NLP pipeline for clusters of related news articles using NLTK, TF-IDF, as well as POS and SRL tagging • Created summarizer using BART and saved for future usage by the Globe and Mail teamClient: Sponsor Circle Ltd. – Systems + Recommender System Update• Restructured and reformatted login configuration files to secure database information and credentials.• Updated the existing recommender system using to correctly utilize new tables created from an update to the database and saved recommendations to a PostgreSQL database hosted on S3• Cleaned codebase by combining and removing redundant files as well as applied general bugfixesClient: Wonderverse/Wonder DAO – Topic Modelling + Visualization• Developed an NLP pipeline in python to filter and clean task data(Languages, NLTK, n-grams, regex)• Created topic modelling visualizations using LDA and BERT to visualize task clusters• Performed EDA on provided task data and created an interactive visual dashboard using Tableau