About Me
I am a researcher in Computer Science at Loyola University Chicago working on trustworthy AI: the engineering challenge of building AI systems that hold up under scrutiny, adversity, and real-world use.
AI is already embedded in banking, hospitals, critical infrastructure, and platforms that hundreds of millions of people use daily. Capability is no longer the hard part. The hard part is reliability under pressure. Most AI research picks one failure mode and studies it. I am interested in the overlap between three of them, because a system that passes one test and fails the others is not something you can put in production and walk away from. My research examines what it takes to build something that actually holds.
Download CVResearch Pillars
Correctness: Formal Methods for AI
Lab: AI4FM Lab · PI: Dr. Mohammed Abuhamad
Formal verification is the practice of proving, mathematically, that a piece of software does what it is supposed to do before anyone runs it. Engineers use TLA+, a formal specification language, to certify distributed systems and cloud infrastructure before those systems are put into service. My work in this area looks at whether large language models can help write and fix those specifications automatically. The goal is to make the process faster and more accessible, and to understand precisely where AI assistance is reliable and where it introduces new risks.
Security: Adversarial Machine Learning
Lab: AISeC Lab · PI: Dr. Mohammed Abuhamad
When a machine learning model is deployed in a sensitive environment, it becomes a target. My work in this area focuses on federated learning, a training approach used in healthcare, finance, and public infrastructure that keeps raw data distributed across devices rather than centralizing it. I study how these systems can be manipulated through poisoning attacks, model inversion, and evasion, and how to build defenses that remain practical at scale. I am also mapping the broader vulnerability surface of large language models, which are now being adopted in consequential settings with limited understanding of how they behave under adversarial conditions.
Safety: Online Safety & Social Computing
Lab: BullyBlocker Lab · PI: Dr. Yasin N. Silva · Supported by NSF SaTC & industry award
My earliest research line deals with harmful content on social platforms: cyberbullying, harassment, and coordinated abuse. Rather than flagging individual messages, I build models that read full conversational threads, which is how harm actually propagates. This work is funded through NSF Secure and Trustworthy Cyberspace awards and an industry research grant. The next phase extends it toward LLM-based systems that can not only identify harmful exchanges but actively intervene in them, with the aim of giving platform operators tools that work at real scale.
The Through Line
Verification without security is fragile. Security without safe behavior is insufficient. Safe behavior without rigorous grounding is hard to reproduce. These three research threads are not independent: each one assumes the others matter, and progress in one without the others leaves the underlying problem unsolved. That is why I work across all three.
Education
Loyola University Chicago, MS in Data Science
University of the People, BS in Computer Science
Microverse, Software Development Program