About
I'm Jamie Fong, an undergraduate at Vanderbilt University pursuing a B.S. in Computer Science and Mathematics. My work sits at the intersection of machine learning and formal methods — I'm interested in what we can actually prove about the systems we deploy, not just what we can measure on a test set.
I do research in Vanderbilt's VeriVITAL lab under Dr. Taylor Johnson. I built the first formal robustness verification benchmark for surgical skill-assessment neural networks, designed a framework for composing and verifying multi-agent behavior trees, implemented in BehaVerify, and developed a pipeline that formally verifies whether a dermoscopy classifier's diagnosis changes with the patient's skin tone.
Alongside research, I'm a teaching assistant for Vanderbilt's data structures course, and I previously worked as a software engineer at Arrowz. I like building things that are both rigorous and genuinely useful.
Education
Vanderbilt University
B.S. in Computer Science and Mathematics
Relevant Coursework
Selected Work
All projects →Benchmark: Formal Verification for Surgical Robots
A reusable neural-network verification benchmark at the intersection of clinical AI and time-series data: surgical skill-assessment networks trained on da Vinci robot kinematics (JIGSAWS), where tools are ranked by the longest time window they can certify.
Python · PyTorch · n2v · α,β-CROWN · ONNX · VNN-LIB
Composing Multi-Agent Behavior Trees
A DSL framework for modeling and verifying multi-agent behavior-tree scenarios through agent-type templates and an expansion-based architecture, implemented in the BehaVerify toolchain.
Python · TextX · nuXmv · BehaVerify · DSL design
Verifying Skin-Tone Fairness of a Dermoscopy Classifier
The first deterministic, post-hoc verification pipeline that certifies whether a skin-lesion classifier's diagnosis stays the same when only the surrounding skin tone is darkened.
Python · PyTorch · α,β-CROWN · n2v · CIELAB · HAM10000
All featured projects came out of my research with Dr. Taylor Johnson at Vanderbilt. Full write-ups and papers are on the Research page.