Ryan Franks Email & Phone Number

Seattle, Wa, Us

Evanston, Il, Us

Greater Chicago Area

• Designed, implemented, published, productionized, and optimized a novel model-agnostic pipeline for identifying and strategically mitigating contributors to model-level disparate impact• Developed heterogeneous graph neural networks using edge features for fraud detection• Built an in-house dataset library and performed meta-learning studies, including empirically investigating the impacts of hyperparameter tuning and missing value imputation

Greater Chicago Area

• Designed, implemented and filed patent on a novel deep-learning framework for datasets with tabular or sequence structures, yielding competitive results in fraud, acquisitions, and AML• Supported a graduate summer intern and a rotating PhD data scientist through their projectsResearch papers: ▫ Wasserstein-based fairness interpretability framework for machine learning models. Machine Learning Journal (2022). ▫ Model-agnostic bias mitigation methods with regressor distribution control for Wasserstein-based fairness metrics. Preprint (2022).

Greater Chicago Area

• Reduced over-fitting risk in AML models by improving the rigor of validation dataset use during adaptive model design• Identified novel AML features using exploratory data analysis in SQL and built models in R

Tampa/St. Petersburg, Florida Area

• Designed and implemented novel outlier detection method with improved accuracy and 40% fewer households flagged• Introduced new metrics for identifying suspicious household viewership behaviors, raising confidence in field operations responses to suspicious households• Streamlined and improved readability of outlier reports for ease-of-use by field operations

Greater Chicago Area

• Created a new analytic thermodynamic model to conceptually capture experimental observations of self-assembling nano-particles and predict trends in nano-particle design• Designed new electron beam lithography patterns for promoting self-assembly and validated them through course-grained molecular dynamics simulations

Evanston, Illinois

• Accelerated LAMMPS molecular dynamics simulation by 5-10x using GPU computing in HOOMD• Gave one hour presentation to research team on how to use HOOMD and migrate from LAMMPS when applicable• Developed new format for molecular dynamics files designed for converting LAMMPS simulations to HOOMD simulations and vice-versa for use in validation

Bachelor Of Science (Bs), Materials Engineering, 3.96

Activities and Societies: Slivka Vice President of Financial Affairs, Undergraduate Research Peer Mentor, Member of Working Group.

Bachelor Of Science - Bs, Applied Mathematics, 3.95

Education record