Applied AI Researcher | NLP & Generative AI
Building intelligent systems at the intersection of research and real-world impact. My work spans large language models, agentic AI systems, and synthetic data generation — with a focus on trustworthy and deployable AI.
About Me
I am a Computer Science PhD from the University of South Florida specializing in the intersection of Artificial Intelligence, Human-Computer Interaction, and Explainable AI. While much of the AI community focuses on performance benchmarks, my research asks a different set of questions — does AI actually help people? Do users understand what its outputs mean? Do they trust it, and how does that trust shape their decisions?
My work sits at an often overlooked frontier: the behavioral and cognitive impact of AI on the people who use it. From studying how scatterplot visualizations influence trust in recommender systems, to discovering that most general users struggle to interpret visualization types beyond basic charts — my research consistently reveals that building powerful AI is only half the challenge. Making it understandable, trustworthy, and accessible to non-technical users is equally critical and far less studied.
Alongside my human-centered research, I build. I architected GenFakeNewsNet, a four-agent LLM orchestration system that produced the most challenging synthetic disinformation benchmark to date. I developed production ML forecasting systems at Amadeus IT Group and healthcare risk models at Florida Blue — bringing the same rigorous thinking I apply to research into real-world deployable systems.
I am driven by a simple belief: AI should work for everyone, not just those in tech. Safe, interpretable, and accessible AI is not a constraint on innovation — it is the standard we should be building toward.
Human-AI Interaction · Explainable AI · Trust & Decision-Making · Data Visualization
IEEE TVCG · Springer ISVC · TopoInVis · 5 peer-reviewed papers
Amadeus IT Group · Florida Blue · Production ML Systems
Tampa Bay Area, Florida, USA
Research & Publications
Applied persistent homology principles to improve force-directed graph layouts, achieving faster convergence and higher quality node-link diagrams. Introduced new algorithms for 0- and 1-dimensional persistent homology features with interactive visualization capabilities.
Investigated the impact of scatterplot visualizations on trust and decision-making in recommender systems through a two-part human-subject experiment. Results show that descriptive accuracy labels (high/medium/low) foster greater trust than numeric accuracy values, with significant implications for human-AI interaction design.
A comprehensive survey examining empirical studies, real-world applications, and open challenges in the use of annotations across information visualization systems, providing a foundation for future research directions.
Developed a comprehensive taxonomy and design space for annotations in information visualization through qualitative analysis of common practices, providing structured guidelines for annotation design in visual analytics systems.
Applied ML-based predictive modeling to forecast friction performance of recycled brake materials, validated experimentally — demonstrating the applicability of machine learning to materials science and engineering challenges.
Projects
Architected a four-agent LLM orchestration system leveraging intent conditioned generation, outlet specific stylometric profiling, and paragraph level linguistic cloning to synthesize 17,400 high fidelity real/fake article pairs grounded in a 30-year, 45K article disinformation corpus. Capped 13 state-of-the-art detection models at F1 of 0.89 across 10 benchmarks. Validated through an 870 participant human study confirming a 52.7% deception rate, nearly triple the sentence replacement baseline (18.9%).
Investigated how model outputs, performance metrics, and visualizations influence user trust in AI-driven recommendations. Designed and implemented recommender systems using tree-based models across healthcare and credit datasets. Applied Bayesian Inference and Structural Equation Modeling to quantify relationships between model design choices and user behavior, demonstrating a 25% improvement in decision consistency when contextual signals accompany model outputs. Published at Springer ISVC 2024.
Developed a Random Forest classifier using multimodal data to detect pain in individuals with 87% accuracy, applied to identify pain signals in soldiers to mitigate risk of drug abuse. Combined emotion classification with clinical decision support objectives.
Built and compared RNN and BERT-based models to predict sentiment of movie reviews, evaluating the performance gap between traditional sequential models and transformer-based architectures on real-world NLP classification tasks.
Experience
Amadeus IT Group · Dallas, TX
Amadeus IT Group · Dallas, TX
Florida Blue · Jacksonville, FL
Teaching & Mentoring
Computer Networks Lab — Instructor of Record
Introduction to Theory of Algorithms — Curriculum Design
Multi-Course Teaching Support
Service & Volunteering
Finance Chair · IEEE VIS Conference
Student Volunteer · IEEE VIS Conference
Skills
Languages
LLM & Generative AI
Machine Learning
Deep Learning
Natural Language Processing
Statistical Analysis
Data Science
MLOps & Deployment
Cloud & Infrastructure
Visualization & Communication
Tools & Platforms
Education
Doctor of Philosophy, Computer Science
University of South Florida · Tampa, FL
Master of Science, Computer Science and Engineering
University of South Florida · Tampa, FL
Bachelor of Technology, Computer Engineering
VR Siddhartha Engineering College · Vijayawada, AP, India
Contact
Whether you're a recruiter, researcher, or collaborator — I'd love to hear from you. Open to industry roles in AI/ML and research collaborations.