- Utilized inverse reinforcement learning to analyze human behavior and interaction with AI from a driving game- Analyzed player data with exploratory data analysis by classifying players into "clusters" based on their behavior archetype

Wrote a training, testing and inference framework for creating bank check datasets and benchmarking algorithms andmodels with Python, AWS SageMaker, and AWS EC2• Classified potentially fraudulent treasury checks with 99.6% accuracy by fine-tuning an open-source multi-modal LLMusing PyTorch• Extracted check and MICR (Magnetic Ink Character Recognition) data with 95% accuracy using OpenCV and OCR(Optical Character Recognition) engines such as AWS Textract and docTR•… Show more Wrote a training, testing and inference framework for creating bank check datasets and benchmarking algorithms andmodels with Python, AWS SageMaker, and AWS EC2• Classified potentially fraudulent treasury checks with 99.6% accuracy by fine-tuning an open-source multi-modal LLMusing PyTorch• Extracted check and MICR (Magnetic Ink Character Recognition) data with 95% accuracy using OpenCV and OCR(Optical Character Recognition) engines such as AWS Textract and docTR• Packaged the LLM and OCR extraction algorithms into a library for deployment into production Show less

Built a marketplace app to sell and list various garden plants with Flutterflow

- Ran particle physics simulations with Geant4 in C++ to simulate muon environments- Developed a pipeline to extract muon flux data to map subsurface density and classify subsurface features using Python

• Building a 3-day predictive power outage quantitative model saving $30mil+ annually across US Walmart stores by improving outage preparedness• Achieved an accuracy of 0.72 area under the ROC curve using Python, Pandas, NumPy, SciKit, and RayTune to develop supervised Machine Learning prediction models (XG Boost, Decision Trees, Random Forests, Gradient Boosted Tree)• Processed 20+ Terabytes of data through an advanced weather forecasting model by utilizing Google Cloud’s compute… Show more • Building a 3-day predictive power outage quantitative model saving $30mil+ annually across US Walmart stores by improving outage preparedness• Achieved an accuracy of 0.72 area under the ROC curve using Python, Pandas, NumPy, SciKit, and RayTune to develop supervised Machine Learning prediction models (XG Boost, Decision Trees, Random Forests, Gradient Boosted Tree)• Processed 20+ Terabytes of data through an advanced weather forecasting model by utilizing Google Cloud’s compute engine through a virtual machine (Ubuntu) and storage buckets to obtain advanced weather data for our outage prediction model Show less

• Wrote the majority of the base code (~90%+) of enemy/entity behavior for our current project, Blipzz

• Built a responsive web widget that displayed facial measurements by processing webcam data using Google’s Mediapipe, rescaling outputs to metric measurements, and rendering them using React JSX and Canvas• Developed an weighted algorithm in Typescript to improve facial mask fit by 40+% by analyzing face measurements• Integrated 3D blend-shapes and B ́ezier curves to further improve the facial fitting algorithm by transforming data with linear algebra and processing them using… Show more • Built a responsive web widget that displayed facial measurements by processing webcam data using Google’s Mediapipe, rescaling outputs to metric measurements, and rendering them using React JSX and Canvas• Developed an weighted algorithm in Typescript to improve facial mask fit by 40+% by analyzing face measurements• Integrated 3D blend-shapes and B ́ezier curves to further improve the facial fitting algorithm by transforming data with linear algebra and processing them using Three.js/WebGL• Researched papers and studies conducted by the National Institute for Occupational Safety and Health to accurately identify key facial features and measurements Show less

Ensured accuracy of account data by utilizing LinkedIn Navigator to update client information