Lead a novel and innovative approach to incorporating mass market promotions with personalized offerings optimizing markdown spend while enhancing the customer experience in highly cross-functional environment. Deliver sales forecasting models predicting impact of mass market promotions suitable for being served in real time that explicitly accounts for the impact of COVID on the brick-and-mortar retail experience. Drive data discovery and dataset construction. Manage relationships with external vendors. Key Achievements: * Developed an XGBoost-based regression model using lagged historical features to predict baseline sales and impact of arbitrary promotions over a three to nine-month timeframe. * Collaborated with external vendors to produce budget allocation recommendations under “sales-heavy,” “margin-heavy,” and “balanced” strategies.

Maintain and extend a custom-built state-of-the-art Kubernetes-backed job scheduler capable of running 10s of jobs with 1000+ tasks across 100s of nodes. Create and implement design documents for quarter-long projects.Key Achievements: * Implemented a system allowing Machine Learning Engineers to combine existing datasets without having to go through the time-consuming process of creating them from scratch. * Developed a system for scraping logs for error messages and automatically creating JIRA tickets for them.

Lead state-of-the-art adoption of counterfactual evaluation techniques of new recommender systems. Partner with engineering and quality management teams in developing tools and techniques to incorporate hard example data mining into a collection of training data based on human assessments without biasing evaluation towards the model used to select the data points. Creating tools to streamline processes and accelerate the onboarding of new projects. Provide the mentoring and training necessary to empower new data scientists and machine learning engineers to perform productively from day one. Develop and teach courses on the skills of communicating effectively with business personnel within a business environment. Conduct presentations and exercises on the concept of adding unit testing to the workflow and writing unit-testable code in support of EDA and model development. Key Achievements: * Designed an automated pipeline building tool by leveraging the Model View Controller (MVC) design pattern to build client applications to be imported into Jupyter notebooks which streamlined and added efficiency to the data pipeline construction process that enables new projects to be onboarded in days as opposed to weeks. * Implemented an offshore labeling methodology to achieve bias-free evaluations consisting of hard examples necessary for the improvement of training.

Developed Jupyter-based data science support tools that improved data scientists ability to perform model development and data science tasks more effectively. Researched and led efforts into state-of-the-art personalization evaluation by leveraging counterfactual analysis techniques. Key Achievements: * Improved model-development velocity by designing a tool for developing and consuming near-production parity recommendations for arbitrary personalization models over arbitrary timeframes in Jupyter. * Designed a framework for clickstream-based evaluations of personalization models that accounted for position bias.

Developed a business-to-business outreach model that assessed each order to identify potential candidates for enrollment into the B2B program. Utilized Scikit-learn and Vertica to design a neural network-like collection of gradient-boosted regression forest ensembles. Designed a standalone Python web service using Flask, Gunicorn, SQL Server, Jenkins and Redis to identify orders that withstood multiple Cyber-5, Wayday, and high traffic sales events.Key Achievements: * Created a B2B outreach model that significantly reduced the number of outgoing phone calls required to generate sales while, at the same time, increasing the new accounts generated. * Implemented a microservice platform for consuming predictions in near-real time.

Developed and maintained a customer-facing PHP code and pipeline of T-SQL stored procedures and Jenkins managed Python scripts to maintain an index of product catalogue for real-time search and faceted retrieval. Key Achievements: * Developed and implemented an evaluation paradigm for keyword search that grouped customer click data into distinct search experiences which became the dominant framework for all customer search KPIs.

Developed a probabilistic framework for evaluation that remained consistent with existing paradigms while still allowing for the use of information-theoretic tools for meta-evaluations. Performed experiments that clearly showed existing information retrieval diversity evaluation metrics were primarily influenced by underlying performance. Key Achievements: * Authored numerous publications including a first author in SIGIR, the flagship conference in information retrieval. * Designed methods for improving evaluation metrics sensitivity to diversity and measures for quantifying it.

Implemented and validated predictive models of document interaction effects on user behavior and created user studies using Python, R and proprietary toolkits to perform contextual and reusable search engine evaluations. Key Achievements: * Developed a contextual evaluation paradigm for keyword search results that resulted in a first author publication at the first-tier WWW conference.