Work
Everything here started as a problem Jason ran into at the workstation. Some turned into papers, some turned into software the UCLA reading room still uses every day. And some of it is the day job itself, so the example reports live here too. Read one before you send a case.
Rebuilt a failing pipeline around DenseNet-121 with CORAL ordinal loss and multi-view fusion: 87.5% accuracy and 0.953 AUC on 3-class severity, validated with 5-fold stratified group cross-validation. The first study to use multi-view CBCT fusion for canine impaction.
A 3D CBCT segmentation model at 80.7% Dice score, trained on open-source data with agentic AI development tools. First place in clinical research, UCLA Research Day 2026.
A full-stack portal for remote CBCT interpretation: custom Cornerstone3D viewer with real-time multi-user viewport sync, Whisper-based dictation with a dental-term correction dictionary, RAG over 568 structured reports for grounded drafting, and automated intake, invoicing, and case distribution.
Co-investigator with UCLA Orthodontics on degenerative change detection across longitudinal CBCT. Built the data backbone for a 3,600+ patient dataset: anonymization with cross-timepoint linking, automated image extraction, and an NLP framework for structuring reports.
Clinical AI specialist on the first 3D deep learning classifier for TMJ osteoarthritis on CBCT (82% accuracy, 81% F1): expert annotation on 800+ scans, methodology input across projects, and radiographic expertise for FDA 510(k) submissions.
Rebuilding the oral radiology reference in collaboration with Dr. Stuart White, bringing a classic teaching resource to the modern web.
Automated report and case distribution used daily by 6 staff, saving 750+ hours a year and taking the department nearly paperless. Plus a web platform that tracks faculty feedback to improve resident reports.
The reads
DE-IDENTIFIED SAMPLES
Samples
If you're considering sending cases, don't take his word for it. Read a report first.
How it works
Short version, so you know what you're getting into.
Findings
AWARDS · PAPERS · POSTERS
UCLA Research Day · Democratizing Dental AI: Agentic Development of a CBCT Segmentation Model
Academy of Oral and Maxillofacial Radiology
UCLA School of Dentistry
UCLA School of Dentistry
Manuscript in preparation
AAOMR Annual Session
UCLA Research Day
AAOMR Annual Session