Researcher.
Engineer. Builder.

Machine Learning & Clinical Research

I'm a machine learning researcher and engineer. I work mostly with video, plus audio and language. My foundation is an MS in image analysis, and most of my work since has been in health — movement disorders like tardive dyskinesia and Parkinson's, and mental-health screening. At Videra Health I lead this work end to end (study design → data → labeling → modeling → validation → deployment). What I care about most is whether results actually hold up: validation, testing for bias, and not trusting findings that might be spurious.

Movement, speech, and neurodegenerative disease: Led development of a production tardive-dyskinesia screening system built from clinician-labeled video, and built validated models for Parkinson's disease, dementia, and a range of movement and voice disorders, including external-dataset validation and longitudinal analysis.

Mental and behavioral health: Built multimodal and NLP-based models for depression, anxiety, PTSD, postpartum depression, bipolar disorder, substance-use outcomes, and emotional distress, including systems for detecting suicidal and crisis-related language in clinical workflows.

Clinical language, speech, and interaction: Developed systems for clinical note generation, key-moment extraction, grade-level and language-complexity scoring, empathy, teach-back, and trust measurement, as well as speaker identification. Designed AI-driven tools that give clinicians feedback on communication quality and patient engagement.

Audio-first and unsupervised modeling: Built audio-only and unsupervised models for respiratory disease, vocal pathology, and neurological change, enabling health assessment from speech and sound alone.

Evaluation, safety, and fairness: I set the validation standards for what we deploy — AUC, Cohen's κ, calibration, and threshold selection — run cross-sectional bias audits across demographic subgroups, and build abstention and quality gates so models withhold low-confidence predictions rather than guess. I'm as interested in a model's failure modes as its headline accuracy.

My specialty is machine learning with video. It's where my computer vision background, clinical research experience, and engineering skills all come together.

Software Engineering Background

I've been writing code for over 20 years, starting as a kid nearly 30 years ago. My engineering background spans backend systems, mobile development, and frontend work. I build production systems from the ground up.

Before Videra Health, I spent nearly 8 years at HireVue. I joined to build Android apps (improved the candidate app from 2.5 to 4.5 stars in 5 months, built the reviewer app end-to-end), then transitioned into data science for about 5 years. On the data science side, I designed, trained, and deployed ML models for NLP, computer vision, and speech recognition. I led innovation work on resume matching and LLM embeddings, and built tools to improve model quality across teams. That gives me close to 10 years of data science experience overall.

My educational foundation is a BS and MS in Computer Science, with my master's work focused on image analysis. Strong background in computer vision, robotics, and scientific computing. This is why I'm effective at building ML systems that process video and audio in clinical settings.

Education

Master of Science, Computer Science — University of Utah
Focus: Image Analysis and Computer Vision

Bachelor of Science, Computer Science — University of Utah