Trust Data Science

Data analyst supporting business strategy and ads team by developing dashboards, tracking key metrics, creating analytical packages, and automating report generation.Key accomplishments:1) Worked on NLP package that displays top articles read in a user segment, designed social network graph package that identifies top influential users in specific cohorts.2) Design dashboards to track advertisement success on an overall client level, as well as an individual program level.3) Automate report generation by creating a flask UI and integrating it with Airflow, a scheduler tool.Tools: Python, SQL, Looker, Flask

TA night and weekend course for 6 months course in data analytics, covering topics from Python, SQL, R, Hadoop, Machine Learning, Javascript and more

Data Scientist Intern who worked with graph algorithms for bot detection and influencer detection.Key accomplishments:1) Performed multivariate and univariate analysis and train random forest for social media bot detection API.2) Constructed social network graph database from data stored on Redshift3) Used network graph algorithms to find communities of influencers as well as communities of botsTools: Python, Neo4j, SQL, Redshift, AWS

Work on projects and learn theory about the entire data pipeline, from data extraction (SQL, web-scraping) to data munging (pandas, numpy) to machine learning algorithms (sklearn). Also worked with big data tools and concepts, such as AWS, Spark, and Map Reduce.https://github.com/allenlu95/Tractor-Case-StudyBuilt a recommender system to recommend hilarious jokes. Includes use of NLP to process joke texts, NMF to explore latent features, and sparse matrix factorization for recommendation. https://github.com/allenlu95/Jokes-RecommenderBuilt a statistical model to predict customer churn for a ride sharing company with above 0.8 f1 score. Methodologies include logistic regression, Random Forest, Boosted Trees, and Ada-boost. https://github.com/allenlu95/Tractor-Case-StudyBuilt regression model to predict tractor prices at an auction. Performed feature engineering, feature selection , and grid-search among Lasso, Random Forest, and boosted models. https://github.com/allenlu95/Tractor-Case-Study

Performed feature selection, and managed database for startup focused on artwork recommendations. Used pytorch implementation of VGG16 for feature selection, and stored features in MongoDB. Improved F1 score by 15% over previous HOG implementation.

Visualized company's website traffic, performed A/B testing for performance on company's productUsed Tableau for visualization, Python, scipy for hypothesis testingStatistical tests led company to further improve on previous model

Created web widgets for PK/PD modeling program ADAPT Used R, ggplot visualization, Shiny for web, ADAPT models Created a more accessible visualization of ADAPT