Get in Touch
Interested in collaborating on research or discussing my work? I'm always open to connecting with fellow researchers, students, and industry professionals.
Research
Embodied AI, Agentic Systems, and Human-Centered Technology
Privacy engineering, agentic evaluation, and behavioral policy, with threads in embodied memory and digital well-being
Overview
My research spans privacy engineering, agentic evaluation, behavioral policy, and human-centered technology, grounded in over fifteen years of software and systems engineering across defense simulation, cloud computing, and later AI operating systems and factory systems. I am especially interested in how agentic scaffolding and policy are measured and operationalized, alongside device-agnostic experience memory that decouples learned knowledge from the hardware it runs on.
A parallel track examines digital well-being, privacy behavior, and social computing: how people navigate disclosure and well-being in digital environments, informed by computational psycholinguistics, large language models, and neuropsychological and sociocultural influences on human-computer interaction. My dissertation examined the trade-offs between privacy and social capital in social media influencer networks, and subsequent work has traced how influencer-driven discourse shapes PII disclosure at scale.
Across both tracks, I favor applied research that produces working artifacts alongside findings. This includes PII-Codex, an open-source Python library for PII detection and severity assessment published in the Journal of Open Source Software, and controlled ablation studies of agent skills and tool use in small language models. The common thread is building capable systems that respect the people who use them, then putting those systems in the open where others can test, extend, and apply them.
Publications
Social Computing
Privacy & Security
Asymmetry, Not Capability: Evaluation Shapes Tool-Augmented PII Detection in Small Language Models
Transactions on Machine Learning Research, 2026 (review in progress)
A controlled ablation of tool-augmented PII detection in small language models showing that measured gains often come from evaluation asymmetries between model and detector outputs, not from stronger detection capability.
PII-Codex: a Python library for PII detection, categorization, and severity assessment
Journal of Open Source Software, 8(86), Article 5402, 2023
PII-Codex is an open-source Python library designed for comprehensive personally identifiable information detection, categorization, and severity assessment in text data, providing researchers and practitioners with tools for privacy-preserving data analysis.
Concentrations
Privacy Engineering
Building tools and methods for detecting, categorizing, and assessing sensitive information in text, including open artifacts such as PII-Codex for privacy-preserving analysis.
Social Computing & Well-Being
Studying disclosure, influencer networks, and adaptive well-being frameworks in digital environments, including how social context shapes privacy behavior online.
Agentic Systems & Evaluation
Measuring how scaffolding, tools, and skills affect small language models, including evaluation asymmetries that can manufacture apparent capability gaps in mixed model-tool pipelines.
Behavioral Policy & Experience Memory
Treating policy as a first-class object for language-model agents, and designing device-agnostic experience memory so learned behavior can move across robots, edge devices, and simulation.
Unveiling Influencer-Driven PII Disclosures in Social Media Discourse
E. J. Rosado, L. Wang, L. P. Dringus, & J. Sun
Digital Transformation and Society, 2026
This study investigates how social media influencers drive personally identifiable information (PII) disclosures in online discourse, examining the patterns and implications of privacy-related content sharing in influencer networks.
The Adaptive PERMA Framework: Reconceptualizing Well-Being as Dynamic and Context-Responsive
E. J. Rosado
Open Science Foundation, 2026
Proposes an adaptive PERMA framework that reconceptualizes well-being as dynamic and context-responsive, extending positive psychology constructs for contemporary digital and social environments.
Privacy vs. Social Capital: Examining Information Disclosure Patterns Within Social Media Influencer Networks
E. J. Rosado
ProQuest Dissertations Publishing, 2024
Doctoral dissertation examining the trade-offs between privacy and social capital in social media influencer networks, analyzing how information disclosure patterns affect network dynamics and individual privacy behaviors.