Carol Roach Email & Phone Number

Bedford, Ma, Us

Currently focusing on developing new ML/AI tools in python, including applications for Unscrambler, ProMV, Multi Case, and next generation software. ● Developed model transfer algorithms using GLS and CCA. Wrote an accompanying model transfer application guide. Trained customer facing representatives and customers on model transfers best practices. Released May 2022.● Developed a model drift/alert algorithm using PLS and XGBoost for classification of failures to improve model sustainability. Developed a novel noise detection algorithm using correlation analysis for this work. Supervised an intern on this project. Completed August 2023.● Developed QA testing for an automated lag detection algorithm. Wrote an accompanying guide on proper application of automated lag detection. Trained customer facing representatives on automated lag detection best practices. Released May 2024.● Developed a novel sample augmentation method similar to MixDA with unique constraints. Tentative release date November 2025.

Naperville, Illinois, Us

As a Senior Data Scientist at Nalco Champion, my work in the Downstream Automation group focused on developing algorithms for instrumentation deployed to the oil & gas industry. ● Developed a classification algorithm in python to supplement regression analyses of crude oil via spectroscopy analysis to determine physical properties related to blending. Additon of classification step resulted in a 10% or greater improvement in regression accuracy. Decreased analysis time, as compared to existing methods, from 2 hours to 2 minutes. Completed November 2020.● As subject matter expert, headed collaboration with third party instrument vendor using MLR to determine presence of treatment chemicals. Included overseeing the completion of contract goals, reporting progress to senior leadership, and testing of new algorithms. Trained field operators on proper wet chemistry and instrument techniques, as well as instrument troubleshooting. Developed statistical analysis of results. Project ongoing at the conclusion of my term.● Conducted statistical analysis using correlation analysis, PCA, and SVD in python to determine contributing variables to heat exchanger performance using historical data. Analysis was used to guide further testing. Project completed March 2020.● Developed excel VBA macros to analyze LC/MS, HPLC, IC, and GC/MS data directly from instruments from various manufacturers. Reduced technician data analysis time from 3 days to 2 hours. Completed July 2019● Various benchtop instrumentation work (UV-Vis, Fluorescence, IR, Raman, SERS, LC-MS, LC-MS-MS, HPLC, IC, GC, GC-MS, GC-MS-MS, GC pyrrolysis) including sample preparation, troubleshooting, maintenance, method development, and data analysis. Wrote new and updated existing SOPs, mentored junior team members on methods and instrumentation, collaborated on paper for NACE.

Austin, Tx, Us

The focus of my role at JP3 was the development of a model maintenance program and creating regression models for real-time prediction of oil & gas properties. ● Developed Model Maintenance (anomaly detection) for real-time measurements using statistical analysis, model statistics, and heuristic knowledge for over 100 instruments and 1000+ models deployed across North America. Required use of SQL and excel VBA. Decreased analysis time from 1 week to 3 hours. Released October 2016.● Created and deployed several Matlab PLS models for real-time prediction of oil & gas properties based on NIR spectroscopic instrument readings. Validated models in python. ● Developed an SVD based training sample selection process in Matlab to increase PLS regression model sustainability. Models created with this method maintained calibration for 4x as long as previous methods. Deployed January 2016.● Developed a custom PCA and PLS classification algorithm in Matlab to identify contamination for a customer. Deployed March 2017.

Ph. D., Analytical Chemistry

B. S., Chemistry