Kathryn Meehan, Ph.D. Email & Phone Number

California, United States

Santa Ana, Ca, Us

• Designed training labels and trained an XGBoost model to identify open mortgage risk• Built a data processing pipeline and engineered new features in Snowflake (SQL) to expand the application of my supervised classifier model to data from an additional business unit.• Performed logistic regression to identify correlations between consumer loan data and risk for internal business partners. Engineered features with SQL and python using data stored on Snowflake. • Leveraged unsupervised learning techniques including clustering and topic modeling algorithms to auto-discover risk categories in title production.

San Francisco, Ca, Us

• Built a python app in three weeks using streamlit to enable companies without ML expertise to compare performance metrics of different object detection algorithms in pytorch• Used Google Cloud Platform Compute Instances to accelerate training and testing

Berkeley, Ca, Us

• I analyzed data from radiation and lidar sensors using python/pandas to benchmark algorithms used to localize radioactive sources, using jupyter notebook for exploratory data analysis• I demonstrated that a new sparse parametric reconstruction algorithm, Additive Point Source Localization (APSL) performs comparably to traditional Maximum Likelihood Expectation Maximization (MLEM) approaches, even in low-count scenarios. I co-presented these results at the IEEE Nuclear Science Symposium and Medical Imaging Conference.• I simulated the angular response of multiple detector systems and collected and analyzed data to validate the simulation which improved the accuracy of our reconstruction algorithms • I built a simulation pipeline in python to streamline submission and post-processing of thousands of simulations on a high performance computing cluster.• Regularly made pull requests to shared git repositories while working on the above code contributions

• Successfully led a proof-of-principle study into being converted to an official, fully-funded program at Brookhaven National Lab• I had full ownership of the data preprocessing and selection steps and had to determine them from scratch since the study involved a novel detector setup.• I conducted a classification analysis where I classified particles produced in the collision by their type.I presented these results at Quark Matter 2017, the largest conference in our field. These results were crucial in convincing the Program Advisory Committee (PAC) at Brookhaven National Lab to approve an official Fixed-Target Program with eight different beam energies for 2018-2019.

I lead discussion sections for Physics 7 (life science students) and Physics 9 (engineers and physics majors). Additionally, I grade exams and host office hours and midterm review sessions. Occasionally I write recommendations for Physics 7 students. For physics 7 I teach 2 sections of 30 students each and for Physics 9 I usually teach 6-7 sections of 100-170 students combined.I have also been a TA for a physics of sports class for non-science majors.I have also lead discussion sections, graded and substitute lectured "Physics of Sports" for non-science majors.

Berkeley, Ca, Us

I participated in a three-week data science workshop organized by U. C. Berkeley’s Graduate Data Science Organization (GDSO) where I worked on a project as part of a team of four, under the guidance of an industry mentor. We built an end-to-end pipeline to analyze biometric data that included data segmentation, feature extraction and training different classification models. I used sci-kit learn for training the models and performing cross-validation to optimize the hyperparameters.The Random Forest and Support Vector machines, with feature selection performed by a genetic algorithm, outperformed the models in the original publication that provided the benchmarking data set.

Berkeley, Ca, Us

Thanks to the DOE SCGSR Award I was able to collaborate with a LBNL scientist, Grazyna Odyniec, for a year to advance my research on identified particle spectra and rapidity densities.

Bryn Mawr, Pa, Us

Advisor: Dr. James BattatI am working as part of the Dark Matter Time Projection Chamber (DMTPC) collaboration and am using ROOT to create a map of the neutron background signal for a directional dark matter detector.

Cambridge, Ma, Us

Principal Investigator: Dr. Richard MilnerCollaboration Co-head: Dr. Peter FisherAdvisor: Dr. Jan BalewskiI worked as part of the DarkLight collaboration to design a detector to detect the A' boson, a carrier of the "dark force". I used ROOT to analyze the detector acceptance and efficiency after making cuts on background processes.

Ph. D., Physics

Master Of Science - Ms, Physics

Bachelor Of Science (B.S.), Physics, Astronomy

Education record