|
Currently, I am a Ph.D. student at Purdue University, advised by Professor Sooyeon Jeong. My research focuses on human-centered AI and multimodal, LLM-driven social agents for personalized mental-health assessment and support. I design agents that build rapport and provide context-aware assistance tailored to individual needs and behaviors, grounded in my background in Graph Neural Networks (GNNs), generative-model evaluation, and Out-of-Distribution (OOD) robustness. I completed my bachelor's degree with Honors in Computer Science and Mathematics at Trinity College in 2021 under the supervision of Professor Ryan Pellico (Mathematics) and Professor Ewa Syta (Computer Science). Email / CV / Bio / Github / Google Scholar |
|
- 07/2026: One paper accepted at the International Conference on Multimodal Interaction (ICMI) Long paper 2026.
- 06/2026: One paper accepted to the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD) Research Track 2026.
- 06/2026: One paper accpeted at the International Conference on Multimodal Interaction (ICMI) Doctoral Consortium 2026
- 2025: Began my HCI research journey with Professor Sooyeon Jeong at Purdue University.
- 10/2024: One paper accepted at Unifying Representations in Neural Models Workshop (UniReps) 2024.
- 10/2024: One paper published at Manufacturing Letters (MFGLET) 41.
- 05/2024: One paper accepted at Uncertainty in Artificial Intelligence (UAI) 2024.
- 03/2024: One paper accepted at North American Manufacturing Research Conference (NAMRC) 52.
- 08/2021: I started my journey with Purdue University
- 05/2021: I graduated Magna Cum Laude from Trinity College.
|
My research develops interactive AI systems for mental-health support, with a focus on how social agents can engage users through natural conversation, reflection, and multimodal sensing. I build LLM-powered agents that help people improve their well-being through CBT-informed and Behavioral Activation principles, while also enabling low-burden, ecologically valid assessment by predicting well-being states from natural interactions and multimodal signals. |
|
I develop low-burden mental-health assessment agents that use open-ended video responses, speech cues, facial behavior, and LLMs to predict validated well-being scores, aiming to make assessment more continuous, natural, and ecologically valid. |
|
I build LLM-powered voice agents for college students with executive-function challenges, combining natural dialogue, CBT-informed lessons, Behavioral Activation reflections, and smart task planning to support time management, organization, and emotional well-being. |
|
Thu Bui, Carola-Bibiane Schönlieb, Bruno Ribeiro, Beatrice Bevilacqua, Moshe Eliasof European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD), Research Track, to appear Thu Bui, Anugunj Naman, Carola-Bibiane Schönlieb, Bruno Ribeiro, Beatrice Bevilacqua, Moshe Eliasof UniReps Workshop, 2024 Developed RAP-GNN, an efficient and performant alternative to end-to-end training in GNNs, reducing runtime by up to 6 times and memory usage by up to 3 times while maintaining or even surpassing performance by using random weights for message passing. |
|
Mai Elkady, Thu Bui, Bruno Ribeiro, David I.Inouye, UAI, 2024 Developed a metric and data splitting method for evaluating Generative Graph Models, enhancing assessment by distinguishing meaningful and novel models from memorization of the training set or production of non-meaningful graphs. |
|
Eunseob Kim, Thu Bui, Junyi Yuan, S Chandra Mouli, Bruno Ribeiro, Raymond A. Yeh, Michael P. Fassnacht, Martin B.G. Jun, NAMRC, 52 and Manufacturing Letters, 42 Introduced a novel approach to detect machining chatter by blending expert knowledge with CNNs. The framework digitizes machine tool and sound data, enhancing chatter event labeling accuracy. With an attention block, the model surpasses baselines in both in-distribution and out-of-distribution testing, achieving impressive accuracies of 96% on known and 94.51% on unknown machine tools. |
|
|
 |
Reviewer
|
|
Teaching assistant (2021-2022 and 2023 - Now)
|
 |
Invited Speaker
|
|
Invited Speaker
|
|
This website uses source code from http://jonbarron.info |