Selected Work

Medical Entity Extraction, mySTR (Leidos)

Designed and built an open-domain medical entity extraction system, productized as a standalone platform with two patents pending. Licensed by the U.S. Defense Health Agency for its My Service Treatment Record (mySTR) program. I owned delivery end to end as full-stack engineer, building the agentic AWS pipeline (OCR, classification, field extraction, validation, and graph indexing) from design through production.

How Leidos enhanced intelligent document processing using agentic AI on AWS

MetaBPL (Leidos)

A dual-pipeline LLM agent for fault detection in manufacturing business logic. Published at AHFE 2025.

AI-Generated Text Detection (Leidos)

A weight-balanced transformer approach to detecting AI-generated text, submitted as Leidos's entry to GenAI Detection Task 3 at COLING 2025.

Authorship Attribution & Privacy,  IARPA HIATUS (RPI)

Proposed attribution via figurative-language features and built the Multi-task Figurative Language Model (MFLM), a fine-tuned RoBERTa detecting multiple FL features. Integrating MFLM embeddings improved on an SBERT baseline by 27% across three datasets. Published in Findings of ACL 2024.

Agenda Detection, IARPA INCAS (RPI)

A textual-entailment approach to agenda detection on social media, applied to the 2022 French presidential elections, reaching 71% accuracy in bilingual French/English identification, with a new bilingual dataset. Published at LREC-COLING 2024.