Joshua Whitehead Email & Phone Number

• Led implementation and design of customer portal, which engaged 98% of users during peak seasons and reduced mean time to perform common tasks by 96% versus prior methods• Prototyped end-to-end predictive workforce management solution, identifying $1MM+ in annual recurring savings opportunities and validating feasibility of automated approach to school scheduling• Architected, developed and oversaw organization’s data infrastructure, including data warehouse and ETL pipelines; led R&D and product design on technical initiatives; and collaborated cross-functionally to scale data, analytics, and automation technologies• Founded internship program and directed inaugural project that reduced onboarding time to digital learning platform by 85% and saved $100,000+/year in administrative cost

• Designed high-performance, deep-learning platform that automated loan origination processes. Awarded “Thomson Reuters Lending Platform Innovator of the Year” Australia, 2018• Developed hybrid deep-learning architecture for document classification, achieving 97% accuracy on classification tasks by leveraging visual and textual features in end-to-end system• Implemented sequence-to-sequence model to parse client documents and populate loan application forms, achieving 94% accuracy on entity recognition tasks across corpus of 200+ recognized document types• Engineered Auto-ML bootstrap for modeling out-of-distribution data, enabling production models to scale from 10’s to 1,000,000’s of training samples in online-learning context• Built scalable, high-throughput APIs for cloud and on-premise operation, serving 100’s of concurrent users and 1000’s of requests/second

Austin, Tx

Sept 2022 - presentMay 2012 - June 2016• Developed analytics tool for poker professionals, inspired by innovations in AI and game-playing• Implemented real-time search algorithm that beat best published exploitability benchmarks by 16%• Scaled up analytics algorithms to handle mixed- and multi-player games with up to 10170 decision points• Developed compressed game representation that reduced space complexity 10x versus reference implementations, enabling application to run on community-standard hardware• Vectorized game tree traversal to take advantage of SIMD instruction sets, reducing wall-clock runtimes up to 32x on modern architectures