Converted a signal machine learning training and annotation tool (OmniSIG) to use MongoDB to store annotations. Technologies: NodeJS, React, MongoDB, Express

Created the backend and frontend for an MVP that uses ElastaAlert to detect anomalies in radio frequency detections and fire off alerts in real-time. Technologies: NodeJS, Express.JS, React, ElastAlert

Developed internal tooling and a web application to allow machine learning researchers and engineers to automate their training and testing of neural net models. Technologies: Python, Dash (Plotly)

Developed various hotfixes and small features that helped elevated customers’ user experiences. Ex: Added dynamically rendered tooltips that could be added to the product through just JSON.

CMSC320 - Introduction to Data Science

Prototyped, developed and trained logistic regression models that would classify whether URLs/web pages visited by employees on the Capital One network were benign or malignant. Tested different text vectorization like TF‐IDF.

Designed and developed a data set search and clustering API with Python, Django, Scikit‐learn and a fine-tuned BERT language model to enable engineering teams at Capital One to access high

quality datasets.

Designed and developed a pipeline to better automate and improve the process of uploading to and utilizing an in house data lake with GPT‐2 and BERT for metadata description auto-completion and classification.

Implemented with Python, SimCAI, an algorithm that models multi‐agent movement and interaction in heterogeneous traffic based on their temporal and spatial states.

Developed a Python script to relabel bounding boxes in a traffic data set.