Tung “Tommy” Tran, Ph.D.
Senior AI Engineer & AI Team Lead, SITE 525 · Ph.D., University of Kentucky
I build generative AI that populates a synthetic internet — the personas, content and networks that U.S. military cyber and information-warfare units train in.
About
I lead the AI team at SITE 525, which builds a synthetic internet for military training: a realistic replica of the online information environment — social media, websites and the personas behind them — used by U.S. Army, special operations and cyber units. My team owns the generative AI behind it, from custom LLMs and content-generation pipelines to agentic systems that drive synthetic personas.
I joined IDS International’s information warfare division in 2020, straight out of my Ph.D., and have stayed with the team as it spun off into its own company, SITE 525. The work has produced three U.S. patents so far — on generating faces from demographic attributes and on simulating social networks and the interactions in them — with more applications pending.
Before that I earned my Ph.D. in Computer Science at the University of Kentucky (2020), advised by Dr. Ramakanth Kavuluru. My dissertation, Deep Neural Architectures for End-to-End Relation Extraction, led to an NIH/NLM R01 grant of over $1.3 million.
Patents
Systems and methods for automatic generation of social media networks and interactions
Granted June 2026 · Second patent from the same 2021 filing as US 12,476,877 B2 · Pending in Europe, Australia and Canada · with D. Shin
Systems and methods for automatic generation of social media networks and interactions
Granted November 2025 · Pending in Europe, Australia and Canada · with D. Shin
Systems and methods for artificial facial image generation conditioned on demographic information
Granted April 2025 · Pending in Europe, Australia and Canada · with D. Shin, J. D. Hoye, M. R. Ehlers
Scenario-based cyber network topology generation system and method
Filed January 2026, pending · with J. D. Hoye, N. Melgar
Selected papers
Attention-Gated Graph Convolutions for Extracting Drug Interaction Information from Drug Labels
T. Tran, R. Kavuluru, and H. Kilicoglu · ACM Transactions on Computing for Healthcare, 2021
BibTeX · tran2020attention@article{tran2020attention, title = {{Attention-Gated Graph Convolutions for Extracting Drug Interaction Information from Drug Labels}}, author = {T. Tran and R. Kavuluru and H. Kilicoglu}, journal = {ACM Transactions on Computing for Healthcare}, year = {2021}, doi = {10.1145/3423209}}
Distant Supervision for Treatment Relation Extraction by Leveraging MeSH Subheadings
T. Tran and R. Kavuluru · Artificial Intelligence in Medicine, 2019
BibTeX · tran2019distant@article{tran2019distant, title = {{Distant Supervision for Treatment Relation Extraction by Leveraging MeSH Subheadings}}, author = {T. Tran and R. Kavuluru}, journal = {Artificial Intelligence in Medicine}, year = {2019}, doi = {10.1016/j.artmed.2019.06.002}}
An End-to-End Deep Learning Architecture for Extracting Protein–Protein Interactions Affected by Genetic Mutations
T. Tran and R. Kavuluru · Database: The Journal of Biological Databases and Curation, 2018
BibTeX · tran2018end@article{tran2018end, title = {{An End-to-End Deep Learning Architecture for Extracting Protein–Protein Interactions Affected by Genetic Mutations}}, author = {T. Tran and R. Kavuluru}, journal = {Database: The Journal of Biological Databases and Curation}, year = {2018}, doi = {10.1093/database/bay092}}
Predicting Mental Conditions Based on “History of Present Illness” in Psychiatric Notes with Deep Neural Networks
T. Tran and R. Kavuluru · Journal of Biomedical Informatics, 2017
BibTeX · tran2017predicting@article{tran2017predicting, title = {{Predicting Mental Conditions Based on "History of Present Illness" in Psychiatric Notes with Deep Neural Networks}}, author = {T. Tran and R. Kavuluru}, journal = {Journal of Biomedical Informatics}, year = {2017}, doi = {10.1016/j.jbi.2017.06.010}}
CV
Last updated December 2022, so it predates my role at SITE 525 and my patents. The summary above is current.