An unsupervised graph-based network visualization of the news providing insight into what is being reported and how it is interconnected, featured by Neo4j. Graphs+ allows for visualizing bias and the differences in coverage between news sources. The project is composed of 3 parts. A scalable crawler collecting a corpus of 250k articles for Doc2Vec ML model training. A pipeline for processing articles including generating document embeddings, knowledge enrichment with DBpedia + ConceptNet, and community detection on an article similarity graph.

Google Assistant Action for exploring the latest news stories, developed for DANAMIC creative agency. The voice assistant makes recommendations for more content and explaining how the articles are relevant by using the underlying knowledge graph. The Google Action showcases some features that can be used to create voice experiences for news and media sites. The content reccomendation engine was deployed as a scalable API using the serverless framework and the Google Action was developed using the Node.js (Jovo) framework.

An XGBoost model for predicting an astronaut’s physical activity based on their biometric data, automating the manual task of logging. The project was selected as a national finalist in the Canadian Space Agency Challenge.