HUMAN-AI INTERACTION · AI SAFETY · ADAPTIVE LEARNING SYSTEMS
Graduate training in kinesiology and computer science.
HCI · Affective Computing · Longitudinal Study
Longitudinal study of relational competence across frontier voice AI: 68 sessions, 83+ hours, four models. Documents a safety–agency inversion: models with stronger safety alignment consistently exhibit lower relational agency. Introduces the HERVAC evaluation framework (six dimensions) and the Nine Circles adversarial stress-testing protocol. Five behavioral measures converge across three methodologically independent evidence streams.
Human Subjects Research · Psychometrics · Pre-Registered
Pre-registered cross-sectional survey (N=55) examining how AI companion use relates to interpersonal communication and help-seeking. Found that AI users show substantially higher willingness to self-disclose to humans (d=1.08), with 57% seeking professional help for issues first discussed with AI. A key dissociation: perceived emotional safety with AI predicts help-seeking but not interpersonal disclosure—suggesting AI functions as behavioral rehearsal, not emotional transfer.
Educational Technology · Adaptive Learning
Educational technology platform: 26 production apps across four certification verticals, 68,000+ exam questions, and 6 learning algorithms grounded in peer-reviewed cognitive science research. Confidence calibration, spaced repetition, misconception detection, and stress inoculation training. Built and maintained by a single researcher orchestrating frontier AI across the full product lifecycle.
Interactive Visualization · Knowledge Graphs · BKT
Adaptive learning platform for AI/ML concepts: 42 concepts across a force-directed knowledge graph, Bayesian Knowledge Tracing with open learner model transparency, Ebbinghaus spaced repetition, and 7 interactive explorations running real computations in-browser. The only platform integrating adaptive mastery tracking, interactive visualization, and knowledge graph navigation in one system.
Computer Vision · Biomechanical Analysis
Biomechanical assessment system using pose estimation and a 100-video dataset with kinesiologist annotations. 30 domain-informed features achieve equivalent classification accuracy to 132 raw pose landmarks with 4× fewer dimensions. A VAE quality model validates against expert ratings (ρ=0.958). Full evaluation: ablation study, quality validation, and latency benchmarks under 1% of a 30fps frame budget.
Wellness Technology · Generative AI · Audio Synthesis
Sleep and dream wellness system with AI-powered journaling and analysis. Gemini-driven dream interpretation with crisis detection and offline queuing, programmatic audio synthesis (binaural beats, brown noise, gamma entrainment) via Web Audio API, voice-to-text capture, and longitudinal pattern tracking across mood, themes, and sleep quality. Privacy-first: all data stays on-device with zero server-side storage.
Grew up in a small Bavarian village, enlisted in the U.S. Air Force at twenty, and spent four years as a C-17 Loadmaster in Operation Enduring Freedom, Operation Iraqi Freedom, and Operation New Dawn. After separating, studied kinesiology at SDSU and CBU, then spent several years living and working remotely across Costa Rica, Indonesia, Peru, and Vietnam.
Each stage pointed toward the same set of questions: how humans learn, how they interact with AI systems, and what happens when those systems get the relationship wrong.