Daniel Breen Email & Phone Number

Irvine, Ca, Us

Irvine, Ca, Us

Seattle, Wa, Us

• Main ML POC for Gift Card WW. In 2022, increased the Gift Card system automation rate by 50% keeping bad debt constant, and in 2023, made system enhancements resulting in net $3 MM annual bad debt reduction for Gift Card keeping automation rate constant. Own several parts of the business on the US team – work with diverse stakeholders interdepartmentally within a large interdisciplinary team to maintain ML automation system in the face of evolving adversaries.• Developed a low latency LLM solution to generate email fraud risk scores, resulting in $1 MM bad debt savings annually. Presented research presentation, providing strategic direction for the team.• Onboarded a system utilizing coupled LSTM model leveraging customer browsing behavioral data, resulting in 10% incremental system automation.• Developed an LLM solution based on data seen by fraud investigators with 5% incremental system automation.• Onboarded new system data sources, including cobrand (CBCC/PLCC) CC signals, resulting in incremental annual GMS savings of $13.8 million and 780k annual bad debt savings.• Developed an autogenerated report for measuring operational workload as measured by ticketing system metrics.• Top 5% among team of 100 scientists for number of resolved tickets. Amazon Star Conqueror Award Recipient and Received accolades for supporting L10 Escalation.• Developed a system of real time monitors and an autogenerated report utilizing a lookalike contrastive order population and shap values to assist scientists in detecting and diagnosing system fraud capture loopholes.• Primary Documentation Author for Departmental LLM initiative, facilitating deeper team understanding and application.• Authored selected Amazon OP1 projects for Gift Card WW, and authored many other project proposals for different planning cycles.

Jersey City, New Jersey, Us

• Designed a T5/Roberta based closed domain question answering pipeline extracting features from a corpus of contract text• Developed an ML system to rank order medical school applicants for a large medical school, enabling rank ordering applicants and predicting 10% of the applicants who would likely fail to acquire a residency.• Developed solr based search engine for a company database. Evaluated against ANN based approaches using txtai and implemented autosuggest functionality.• Led a team of junior scientists to build an E-sports influencer recommender system. Developed data pipeline including web scrapers with Scrapy/Splash to asynchronously parse Twitter, Instagram, Social Blade, and Twitch profiles to build SQL database

La Jolla, Ca, Us

• Used and developed methods of parameter estimation in dynamical systems using analogies with the path integral in quantum mechanics. Learned about topics in nonlinear programming including ill conditioning, newton’s method, and sampling methods such as the metropolis hastings algorithm.• Identified differences between diseased and healthy neurons consistent with clinical observations of Alzheimer’s patients using estimated parameters in dynamical models of neurons. • Established a mapping between configurable and true parameter values on neuromorphic VLSI chip, leading to emulation of a biological neuron on VLSI hardware for the first time.

Doctor Of Philosophy (Ph.D.), Physics

Master’S Degree, Physics

Bachelor’S Degree, Physics