Helping companies implement AI products that increase the value of their business.

In my current role, I lead the development of data science projects that require data science expertise and business acumen, with a focus on Marketing Cloud and Data Cloud, two strategic Salesforce priorities. As an example, I led the development of a self-learning, real-time AI decisioning service that personalizes web content. This service, built on a contextual bandit framework, integrates XGBoost and Deep Learning models, selected through an innovative automatic tournament based on off-policy evaluation to enhance clicks and conversions. I crafted a bespoke automatic training infrastructure and devised a system for efficient test data collection and management. Furthermore, I significantly improved the accuracy of an existing recommender system delivering 470 billion recommendations annually by implementing an automatic hyper-parameter tuning system and expanding the embedding size for each client, markedly boosting the system's performance. My work also extended to designing and implementing Salesforce's next-generation Large Language Model (LLM) based recommender for the Genie/Data Cloud platform, currently in development, setting new benchmarks for personalized recommendation services.

As a Senior Data Scientist at Evergage, I spearheaded the development of the Contextual Bandit product, optimizing real-time web personalization through AI. This involved leading the project from conception to implementation, including model design, real-time serving pipelines, training infrastructure, and data collection/storage systems. My role was pivotal in utilizing response data from end-users to continuously refine our decision-making models, ensuring that promotional materials were optimized for engagement. Additionally, I contributed to marketing materials, client interactions, and the overall strategy, driving forward the company's data science capabilities before its acquisition by Salesforce.

At Divert, my role involved leveraging data to enhance recycling efforts for large grocery chains, focusing on cardboard and organic waste. I managed to double data capture efficiency by establishing and refining collection metrics. My efforts in improving data quality through filtering and interpolation led to more accurate analysis, enabling the identification of recycling policy non-compliance. By applying machine learning to calibrate weight detection from pressure sensor waveforms and integrating environmental insights with behavioral economics, I successfully increased recycling rates by 20%.

I architected a therapist matching system at Outcome Referrals, enhancing patient outcomes in behavioral health clinics through data-driven assignments. This system, initially developed in R, was later progressed using Scala and Spark. My work also involved analyzing therapist performance to co-author a paper on identifying their strengths and weaknesses, utilizing machine learning and hierarchical linear modeling. Additionally, I developed a suite of tools in R for custom reporting on behavioral health, aiding in the research and operational needs of state child welfare systems.

In my role at Atmospheric Environmental Research, I was instrumental in developing the Inspect & Analyze application, a Java-based tool for environmental data analysis, from its preliminary design to final delivery. My leadership as the technical lead involved direct interaction with key stakeholders, managing development teams, and integrating innovative features for enhanced data visualization. I also contributed to internal R&D projects, preparing for future satellite missions through software development and data analysis.

At Fidelity, I developed a regression testing harness for the company’s in-house security trading engine, enhancing testing efficiency and reliability. My responsibilities encompassed database schema design, object creation for test case descriptions, and integrating with the trading engine, using technologies such as Hibernate, SQL, Maven, and Java. Additionally, I managed the Hudson (now Jenkins) setup for our project, facilitating continuous integration and deployment processes.

Software testing and verification, miscellaneous intern tasks.