LAI^JASON ORAL & MAXILLOFACIAL RADIOLOGY UCLA-TRAINED

JASONL.AI SERIES: PORTFOLIO ACQ: 2026

X: 0000 Y: 0000 SIGNAL: LIVE

SCROLL FOR PROJECTS ▾

Oral & Maxillofacial Radiologist · UCLA-Trained

Dr. Jason Lai reads the scans,
then builds the software that reads them with him.

Dr. Jason Lai is an oral and maxillofacial radiologist, UCLA-trained. More than 2,000 CBCT studies read, and a growing stack of imaging models and clinical tools he built because he wanted them at the workstation.

See the work → Get in touch

WW: 256 WL: 128

Drag to window · click to open

01 / About

The radiologist in question.

An oral & maxillofacial radiologist, UCLA-trained, 2,000+ CBCT reads deep, and building the imaging software the reading room he trained in still uses.

He reads the scans, then builds the software that reads them with him. The research and the clinic feed each other.

Position: Oral & Maxillofacial Radiologist · UCLA-trained
Focus: Deep learning on CBCT
Toolbox: PyTorch · Cornerstone3D · React

Full clinical history →

02 / Collaborate

Where Jason can help

Start a conversation →

Radiology reads

Interpretation of CBCT, panoramic, and intraoral imaging with clear, clinically useful reports.

AI research collaboration

Clinical grounding for imaging AI projects, from dataset design and labeling protocols to model evaluation.

Continuing education

Lectures and courses on oral radiology and what AI actually changes about practice.

Expert opinion

Records review and expert witness work where imaging findings are the question.

03 / Findings to date

By the numbers

PULLED FROM THE WORKSTATION

0CBCT studies read

0AUC · canine impaction model

0staff hours saved / year

1stUCLA Research Day 2026

04 / Selected work

Featured projects

All work →

P–001 ResearchLead developerCBCT

Multi-View Deep Learning for Classification of Maxillary Canine Impaction Severity Using CBCT-Derived Images

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. The first study to use multi-view CBCT fusion for canine impaction.

Read the study → In preparation · 2026

P–002 1st placeResearch

Democratizing Dental AI: CBCT Segmentation

A 3D segmentation model at 80.7% Dice, built with open data and agentic AI tools. First place in clinical research at UCLA Research Day 2026.

Project notes →

P–003 SoftwareFull-stack

Teleradiology Platform

A full-stack portal for remote CBCT reads: custom Cornerstone3D viewer with live multi-user sync, Whisper voice dictation tuned for dental terms, and RAG-grounded reporting.

See the build →

05 / From the reading room