Applied AI, with people in the loop
Undergraduate Researcher in ASU's CoRAL Lab. Both of the threads below sit at the intersection of NLP/applied AI and a genuinely human problem — how people work together, and how to protect people who don't have technical defenses of their own.
Behavioral Team Formation
CoRAL Lab, Arizona State University · Undergraduate Researcher
Automated or algorithmically-assisted team formation — in education, hackathons, or workplace settings — typically optimizes for skills or availability, but tends to ignore behavioral compatibility signals that strongly predict whether a team actually functions well together. This project investigates behavioral signals as an input to automated team-formation algorithms.
Findings from this work are informing a paper in progress.
On-device scam detection (SentinelEdge)
Built out of a 1st-place hackathon project — full case study →
A paper describing SentinelEdge's approach to real-time, privacy-preserving scam detection — a compressed, fine-tuned small language model performing transcript analysis entirely on-device, with federated learning updates and zero data leaving the device — has been accepted to the EMNLP 2026 System Demonstrations Track. I'll be presenting the work at EMNLP 2026 in Budapest, Hungary.
Abstract and demo link will be added closer to the conference.
Research interests
Applied NLP, on-device/privacy-preserving ML, and the human-behavior side of team and system design — how technical constraints (privacy, compute) can be the design driver rather than an afterthought.
Future directions
Extending the behavioral-signal approach beyond team formation to other collaborative-AI settings, and continuing to explore where on-device inference changes what's possible for privacy-sensitive applications.