Jeff Wen

• Collaboratively developed an advertisement bidding system that uses Bayesian methods to optimize daily bids and budgets for thousands of campaigns resulting in a ~13% increase in margin
• Architected and implemented a brand suitability classifier for YouTube videos that uses natural language processing (NLP) techniques to achieve ~0.92 AUC
• Conducted user research and programmed a chatbot using API.AI that summarizes and provides on-demand highlights on campaign performance
• Managed and updated Docker containers to enable quick iteration and seamless development between engineering and data science teams
• Partnered with business and engineering leadership to define and track data science goals

• Created interactive RShiny applications that queried from Mongo and SQL databases to map and calculate statistics regarding customer density
• Used unsupervised learning algorithms (DBSCAN) to identify clusters of customers who are underserved by current sales and service facilities
• Wrote custom Python packages to help abstract away command line permissions handling for remotely hosted applications
• Projected demand and capacity constraints for existing and planned service locations given expected customer distribution
• Implemented regression and decision tree techniques (in R) to identify key levers in customer satisfaction

• Built a web application that uses computer vision and image clustering techniques to create a visual food search experience for users
• Analyzed presidential speeches using latent dirichlet allocation (LDA) to extract topics and visualized changes through time using D3.js
• Scraped movie data using BeautifulSoup and Selenium to predict total adjusted domestic gross of recently released movies
• Classified different types of job seekers using a majority vote classifer and presented the details in an interactive D3.js dashboard
• Aggregated NYC MTA subway station foot traffic data and mapped the results using CartoDB
• Metis is a 12 week immersive Data Science bootcamp covering topics in statistics, machine learning, programming, communication, and design

• Compared the accuracy of different classification models (decision tree, random forest, SVM, Naive Bayes) to evaluate prescribers' affinity toward in-person sales channels
• Built regression models to predict enrollment in clinical trials for a $117 billion biotechnology company
• Gathered requirements for a supply chain data warehouse and interfaced with clients to create Excel mockup reports
• Designed and coordinated with off-shore developers to create business intelligence dashboards (in Tibco Spotfire)

• Built an Excel based database reporting tool for a $93 billion pharmaceutical client
• Programmed an analytics tool to determine promotion results for a 200+ person sales team

• Conducted pattern analysis on 5,000+ VC firms to understand investment decisions in the clean technology industry

• Built and analyzed Excel datasets from SEC Filings (10-K, DEF14A, 10-12B) on 300+ Fortune 500 corporations in Professor Emilie Feldman's research group

• Competed and placed in the top 6% (87/1,388 entries) in the What's Cooking? competition to classify cuisines based on a list of ingredients
• Trained and evaluated models (logistic regression, random forest, SVM) using cross-validation and grid search on parameters; winning model was an average of the different models

• Self-taught Python to analyze Fortune 500 company annual statements to identify sentiment around corporate sustainability and environmental topics
• Used NLTK to extract sentences related to environmental topics and clustered by industry to look at cross industry differences

linkedin:
linkedin.com/in/wenjeff
github:
github.com/jeffwen
website:
jeffwen.com

proficient:
Python, R, SQL,
MongoDB, Git
learning:
Processing, D3.js,
JavaScript, HTML,
CSS, L^aT_eX

fluent:
english
cantonese
limited:
mandarin