My cybersecurity threat detection capstone used supervised and unsupervised learning to detect and classify network threats in imbalanced datasets. Ensemble tree-based models achieved an F1 score of 0.97 and recall of 0.95.
I also built and fine-tuned a DistilBERT model to classify more than 200K news articles across 41 categories. It achieved a test accuracy of about 0.67 and a macro F1 of 0.59.
I'm building a multi-agent application with LangGraph and CrewAI, implementing RAG for context-aware and summarization tasks. My background in molecular and cellular biology includes research projects involving CRISPR-Cas9, optogenetics, and lung adenocarcinoma.
