• Implemented and finetuned machine learning anomaly detection methods to enhance Consumer and Commercial card risk detection for one of America’s largest banks. Generated 2,700 alerts for potential suspicious activity and led a team of 5 investigators to investigate alerts.• Developed 216 NLP features to enhance one of America’s largest bank’s P2P Suspicious Activity detection model, resulting in a 4% increase in average precision. Optimized NLP pipeline to improve efficiency by 80% using Snowflake queries and Python scripts.• Led user acceptance testing for new tools at one of the largest fintech companies in the US, identifying and resolving critical transaction data mismatches. Enhanced workflow efficiency by 50% through the development of Excel macros and implemented automated analysis of RedFlag Indicators to auto-escalate the top 5 typologies.

• Engineered a comprehensive data preprocessing pipeline in R that enhances data quality and expedites the ETL process.• Conducted multi-faceted growth curve analyses, producing a range of visualizations including scatter plots and line graphs via ggpubr, to interpret the impact of different growth mediums on growth metrics for E. Coli.• Contributed to key research findings by enabling more accurate and timely analyses, thus accelerating the laboratory’s research capabilities and impact.

• Developed and implemented a new set of functionality in EconML, an automatic causal inference package created by the ALICE team, that simplifies hyperparameter tuning and model selection for non-technical users.• Updated EconML estimators that perform cross-fitting to take a model specification, resulting in more accurate default values and improved performance.• Implemented a new approach to model selection by moving the process before cross-fitting, resulting in the selection of the best model and reducing computation time.

• Managed 5 groups of 6 students each to solve classification problems such as stroke and heart disease prediction and deployed model and analysis to a website.• Taught data analysis fundamentals, visualization techniques, and machine learning methods to groups of students of varying technical levels.• Helped iterate over lessons and create notebooks to create a well-paced and balanced curriculum.

• DoseMate: A Real-world Evaluation of Machine Learning Classification of Pill Taking Using Wrist-worn Motion Sensors published at CHIL ConferenceApplied feature-based machine learning (XGBoost, random forest) and deep learning models/techniques (Transformer, CNN, LSTM, and transfer learning) to classify the unique signal of a pill intake, achieving a 95% F1 score.• Utilized signal processing techniques to engineer different features to best categorize a pill intake by preprocessing accelerometer and gyroscope data, transforming the data, and extracting relevant features.• Data was collected via wrist-watch in IRB study, analyzed through SQL, and Python was used to create data visualizations

Reconciled financial accounts in campus retail systems using Microsoft Excel. Implemented changes and repaired campus cash registers using Oracle, CSGOLD, and Symphony. Provided support and troubleshooting via phone