Tia Miceli Email & Phone Number

Northbrook, Il

* Claims and Fraud * Building gradient boosted machine learning predictive models to recommend certain insurance claims be reviewed for fraudulent activity. Aiming to surpass accuracy of manual recommendations. * Productionalizing this model to run on new data automatically, send referrals to investigators, and save records for future legal reference. * Developed new modeling features based on constraints from the data. * Compared multiple models to find the best one without overfitting by using test statistics. Investigated and verified the model predictions were reasonable. * Investigated and chose best segmentation of historical data into train, test, and validate samples to ensured there is no bias our models. * Collaborated with business partners in Claims management to leverage their domain knowledge about the data.* Agency Compensation * Built predictive generalized linear model to improve estimate of bonus payouts to agencies as well as predict which agencies will close in the next year. * Created novel features indicative of agency success by learning agency pay structure and bonus structure. Investigated these across various regions. * Revised models of how agents react to different incentives based on natural experiments so that we could identify and engineer incentives for agents. * Investigated clustering agencies by geographical, economic, and business performance parameters in order to identify best markets and incentives. * Introduced coding best practices to team members. Led development of our own standards including proper usage of functions, branches, pair-coding, code reviews, and testing before pushing to a git repository. * Learned SQL to create useable modeling files from raw Oracle tables. * Partnered with another team to combine domain knowledge on agency survey data. * Collaborated with our business partners in Protection Finance to teach them R as a powerful alternative to Excel.

Fermilab

* Analyzed very rare neutrino interactions within hundreds of terabytes of background data using boosted decision trees using distributed computing. Best feature could also be confused with background interactions.* Improved experiment’s capability to discern signal and background by upgrading detector timing resolution. Pushed for the implementation of multidimensional feedback programs to tune the location of the particle accelerator beam.* Co-led a team of 6 people to characterize background noise in a neutrino beam. * Responsible for identifying human and material resources for projects, designed experimental analyses to fulfill our group’s deliverables to the wider experimental collaboration.* Talented technical communicator: 8 invited lectures, 3 popular articles, and 4+ YouTube videos & podcasts.* Self-taught machine learning concepts and techniques.

* Statistical analysis of large datasets using distributed computing (>10 TB).* Statistically analyzed results of multidimensional data and simulation to determine uncertainties using negative log likelihood using Bayesian weights and published the best experimental limits on a theory of extra-dimensions.* Created new features using principal component analysis to discriminate signal from background. * Determined the signal efficiency and false-positive rate of each algorithm.* Self-taught in automated data acquisition system programming.* Instructed algebra-based laboratory physics classes for a year and graded labs, quizzes, and exams.* Trained through coursework and research in how to breakdown complex problems, ask the difficult and important questions to ensure we are doing things right.* Conditioned to learn new techniques, skills, concepts, in order to successfully complete projects.

Doctor Of Philosophy (Phd), Experimental High Energy Particle Physics

Master Of Science (Ms), Physics, 3.78

Bachelor Of Science (Bs), Physics, 3.89

Degree denoted with "very high honors".