Examples of our work

Optimized Marketing Spend based on insights from path data:  Developed model to analyze clickstream data applying bayesian multi-resolution spatial analysis depicting consumer movements across geographic areas to optimize marketing spend and targeting.

Re-designed internal analytic models for client extracting actionable insights from multiple large high dimensional data sets using machine learning algorithms and spectral methods for data analytics.  Reduced analytic processing time from days to an hour.

Created Marketing Scenario Simulations:  Created marketing insight models to perform marketing simulation scenarios using cross-sectional transaction data summaries helping marketing managers target and engage based on specific insights into customer behavior obtained through performance metrics of targeted messaging.

 

Lowered Cost of Processing Transactions while Increasing Transaction Volumes:  Rewrote transaction modules using Queueing Theory to dramatically increase performance, but also consolidate several diverging code paths into a single service module. With parallel processing, optimized database access, and elimination of unnecessary remote calls, the throughput increased 200X per minute. 

Designed Big Data computing platform: Designed distributing computing platform with ability to train thousands of computers on a single task, to analyze the effectiveness of a company's online applications and the behavior of its users capturing over 100 terabytes of data. 

Designed Machine Learning platform: Used for predictive analytics of social media content. 

Built Solutions using Natural Language Processing: Built document classifiers into Yahoo directory categories and wrote implementations of Naive Bayes / clustering algorithms for text recommendation systems.

Developed New Products: Architected a brand new web-based product, helped hire and coached the development team, and guided the implementation. The product is a powerful multi-threaded fee calculation engine, employing several open source tools including Spring, Hibernate, Apache Commons, Joda, Quartz, log4j, JFlat, et cetera.

Created Predictive Models to Increase Customer Engagement and Lifetime Value:  Developed model to predict future purchasing patterns for a customer base. Created beta-geometric/beta-Bernoulli (BG/BB) model to capture underlying behavioral processes used to predict the future behavior of a customer to increase the customer lifetime value. Results expressed in a format easily digestible by decision makers.

Created Algorithms for Next-Generation Platform: Advised subject matter experts on adapting algorithms in diverse problem domains (eg, image/signal processing, gene sequence matching, stochastic financial modeling, cryptography, language translation) to massively parallel architecture to improve predictive capabilities. 

Improved Anomaly Detection: Designed multi-signal analysis capability to internal anomaly detection libraries and pipelines. Presented introduction and analysis of variant signal decomposition methods to anomaly detection group. Identified highest-value signals for automated analysis and alerting.

Improved Data Quality combined with Improved Turnaround Times:  Evaluated and redesigned system starting with data ingest and running all the way to data model building —optimizing system at every point to drive down cost.  Reduced data capture errors, raised efficiency of data delivery by automating routine tasks, increased processing speed, and created usable insights in a flexible format that can be easily interpreted by decision-makers.  Data-to-Insight time turnaround time has been improved by 80% while reducing the involvement of developers and system engineers.