• Leading backend-development and deployment of data retrieval system for querying academic information • Selected for UVA Darden School of Business iLab Incubator program

• Implemented an LLM-based RAG application using LangChain, Hugging Face, and Ollama, embedding system log entries in a vector database and employing a self-query retriever to generate SQL queries based on user input for precise log filtering and retrieval• Developed a network preprocessing pipeline to process pcap files, utilizing Tshark for feature extraction and engineering, reducing manual analysis time by 40% and integrating Snort and Suricata alerts for labeling packets in feature store• Optimized and tuned ARIMA time series models for CPU utilization on AWS EC2 instances, generating anomaly alerts to correlate with IDS security events

• Developed a long-short-term-memory (LSTM) based GAN using reinforcement to generate novel packet-level training data for intrusion detection system models (IDS), safeguarding against adversarial machine learning attacks• Designed a Kali Linux virtual environment, integrating Wireshark and Metasploit to generate and capture malicious network traffic, producing a dataset that trained an XGBoost binary classification model with 99.48% accuracy in identifying malicious packets• Directed an internship capstone team in creating a custom LLM-driven framework, integrated with interactive Tableau dashboards, to enhance data collection and user feedback for participants in internal cybersecurity Capture the Flag events

• Leading team of UVA researchers in adapting 2D-UNET codebase for pancreatic islet segmentation and quantification

• Awarded the Harrison Undergraduate Research Award for implementing 2D-UNET CNN model for segmenting and quantifying atherosclerotic plaque burden in aorta of mice on histological sections with: R2 = 0.944, correlation = 0.976 (pending publication)• Deployed model for rapid measurement of atherosclerotic plaque sizes in mice, replacing the need for manual measurement of lesion sizes, reducing measurement time by 98.88%

• Analyzed AI-based oncology value-based care models and recommended the strongest potential investment opportunities by constructing a care model spectrum tool to evaluate the condition-focused care common components and identifying the key levers for value creation in the oncology care market• Created an automated web scraping program in Python using beautiful soup library to identify and surface individuals and contact information for 2,000+ potential limited partner investors

• Edited and updated post-operative informational documents that outlined restrictions and therapeutic guidelines for patients recovering from arthroscopic procedures. • Created and designed educational pamphlets for recovering patients that informed on anatomy, physiology, injury prevention, diagnosis, recovery, and symptoms.• Observed several hip, knee, and shoulder arthroscopic procedures, along with clinics and diagnoses.