William Calhoun Email & Phone Number

New York City Metropolitan Area

San Francisco, California, Us

Stockholm, Stockholm County, Se

Currently manages a hybrid team of 8 across Data Science and ML Engineering to create data products (mainly production ML models and data pipelines) that power Spotify’s Concert/Merch personalization and discovery products. I am responsible for driving high impact projects, setting technical strategy for the team's ML architecture, and developing more junior members of the team. Personalization example: A DNN powered recommender system that takes into account a user’s behavior, a concert’s lifecycle, and ticket prices to optimize marketing opportunities in the app and via email. This system controls hundreds of millions of user impressions and millions of ticket purchases through spotify.Forecasting example: A multi-tier model system for predicting how many tickets would be sold for a ‘hypothetical’ concert based on aggregated user data and econometric signals.Also have a strong track record of quickly iterating POCs in the ML space and seeing them through to full production use.

Stockholm, Stockholm County, Se

Stockholm, Stockholm County, Se

New York, New York, Us

Principal Data Scientist on Machine Learning Projects involving Grand Theft Auto V and Red Dead Redemption 2Cheater Identification: Anomaly Detection Techniques with Deep Learning, Cheater Classification Model Implemented with h2o.ai and keras and tensorflowPlayer Behavior Predictive Analytics: Player Conversion, Player Churn, Online Economy ML Models -- all live in productionPlayer Behavior Research: New Content Research, Player Clustering/Archetype Analysis, Deep Variable Extraction

New York, Us

Implemented A Model to Infer Home Location for ~10 MM People Based on Credit Card Transactions. Validated model results to Under 1 Miles Avg Across the Entire US. Over 30 Billion Rows of Data.Deep Data Discovery on Affinities between different Retailers in Credit Card Data. Network Structure Analysis, Clustering Company Relationships, Modeling User Churn and Growth, and investigating causal linksDesigned and implemented models to predict promotion performance on item inventory (All Items, All Stores), predict probability of flu shot based on historical purchase trends—using large Pharma retail data (>50 Mil Customers)

New York, Us

Data Science application development: Created Tools designed to help audit new data sets and preform various analysis in a UI friendly manner on large data sets.Created a "Fuzzy Linking" Statistical Match Tool and applied it to matching specific retail data to credit card data to generate customer profiles and assess Customer Lifetime ValueDesigned models and procedures to predict promotion performance on large pharmacy/drugstore data so as to accurately predict the effect on inventory levels at all stores for all itemsHelped add basic Time Series functionality to the 1010systemCreated Sales Optimized Product Generic for any Retailer based on Credit Card DataHelped Geo Tag Credit Card Transactions based on descriptions and tag Building Permit Descriptions to Companies

New York, Us

Creating a Dynamic Linked Library of open source Support Vector Machine (SVM) algorithms.Applying these algorithms efficiently to massive amount of data on the 1010 system.Ultimately created a customer facing function to run SVM analysis on " Big Data"

Toronto, On, Ca

Research project into the affect of "sentiment" on Foreign Exchange prices. Using novel sentiment metrics and over 25 currency pairs.A working paper is still in progress on our findings.