Building the centralized DS org from scratch for the company. Besides full-stack DS, hired the company's initial MLE and LLM talents. Deliver DS and ML solutions to inform and drive company decisions, strategies, and actions across Engineering, Product, Go-To-Market, Finance, etc. Current focus areas include company priority metrics, product analytics, forecasting, growth, pricing & commercialization, anomaly & abuse detection, recommendation, and experimentation, GenAI, search and assistant.

Lead a centralized and horizontal Data Science and Applied Machine Learning org for Disney+, Hulu and ESPN+. Part of the executive team responsible for launching Disney+ across multiple regions and countries. Built the org from scratch, now with teams in NYC, SF Bay Area, LA and Seattle.Current focus areas are engagement and retention predictions (e.g., churn, up-sell, customer value, insights), forecasting, audience segmentation, anomaly detection, anti-fraud, personalization, customer service and chatbot, experimentation platform and A/B testing.

Back to Apple to build and lead a new horizontal and cross-functional data science org with teams in UK and US focusing on product data science, machine learning engineering, applied research, and data investigation & analytics for AMP (Apple Media Products), covering products such as App Store, Apple TV / TV+, Apple Music, Apple Arcade, Podcast, and iTunes Store.Some major focus areas include insight models, anomaly detection, machine learning platform, chart ranking and popularity scoring, spam and fraud detection.

• Apr 2017 - Oct 2017: Director of Data Science, Consumer and Merchant Business• Mar 2016 - Apr 2017: Director of Data Science, Consumer BusinessLed PayPal centralized consumer and merchant business data science teams applying machine learning, NLP and AI techniques to create predictive models and personal recommendation engines for global (NA, EMEA, APAC, LATAM) marketing, consumer and merchant product, customer service, and risk at PayPal. Some major focus areas include customer value prediction, churn and reactivation prediction, next best action models (e.g., recommend next product / shopping category / merchant), personalization models for product and customer experience, response and propensity models for marketing, SEM keyword bidding optimization, customer contact/intent prediction, and chat-bot.

- Data mining and modeling, search relevance for App Store, Apple Music, Siri, Apple TV, iTunes Music.- Led data mining and search relevance efforts for CJK (China, Japan and Korea) and other international markets, e.g., tokenization, query understanding and rewrite, spam detection, and machine learned ranking.- Features/products shipped to production internationally with record-breaking improvements on both online and offline metrics.

R&D on relevance of Bing captions (i.e., search result summaries, mainly consisting of title, snippet and displayed URL):- Led machine learned title and snippet selection- Led Bing caption title projects: title generation, new title sources, title ranking/selection, title synthesis from different sources, metrics design for measuring title quality- Search log and Web index mining using MapReduce to identify patterns and key insights of captionsReceived Microsoft Gold Star Award

R&D on semantics, knowledge resource, and search relevance ranking for building a next-generation natural language search engine.Led machine learned ranking: Designed, implemented, and deployed a machine learned search result ranking system for the natural language search engine, including system design, feature engineering, training and testing data generation, modeling, feature selection, and model deployment on production. New model significantly outperformed previous rule-based production ranking system as well as our competitors.Powerset was acquired by Microsoft in 2008.

• Research areas: NLP, Machine Learning, Knowledge Representation, and AI • Ph.D. advisor: Dr. Jerry R. Hobbs (AAAI Fellow, ACL Lifetime Achievement Award in 2013)• OWL-Time (formerly DAML-Time): Developed an ontology of temporal concepts in first-order logic and OWL Web Ontology Language for the Semantic Web and natural language applications.• TARSQI (Temporal Awareness and Reasoning Systems for Question Interpretation): Constructed an annotated corpus of typical durations of events from news articles; applied machine learning techniques to the data to automatically extract implicit and vague event durations from texts. • CALO (Cognitive Assistant that Learns and Organizes): Developed ontologies for different domains.

Skill Affinities: Designed and implemented a semantic analyzer for computing similarities between skill descriptions in natural language, which can be embedded in a search tool for matching available employees to open job positions to allow searches to be expanded to related skills and return more potential matches, instead of only returning exact matches.