I build virtual agents with rich human-like behaviors.
I'm a PhD candidate at the Khoury College of Computer Sciences at Northeastern University, where I work with Prof. Stacy Marsella in the CESAR Lab. My research focuses on generating multimodal behavior for embodied virtual agents, generating rich nonverbal behaviors (gaze, gestures, facial expresion, etc) that are appropriate to the role of the virtual agent and the context of the interaction between the agent and the human .
Recently I've been using large language models as semantic selectors, selecting appropriate behaviors that the agent should perform during an interaction with the user.
Outside of work, I like to dance, embodying emotions by taking forms and motions.
I'm always glad to hear from people working on gestures, nonverbal behavior, or designing and deploying virtual agents in different applications.
Research interests
- Multimodal behavior generation
- Embodied conversational agents
- Co-speech gesture
- Nonverbal communication
- Large language models
- Human-agent interaction
Publications
* equal contributionLarge Language Models for Virtual Human Gesture Selection
Towards an Intelligent Augmented Reality Tutoring System for Precision Manufacturing
Investigating Large Language Models' Perception of Emotion Using Appraisal Theory
Large Language Models in Textual Analysis for Gesture Selection
Talks
Large Language Models for Virtual Human Gesture Selection
Towards an Intelligent Augmented Reality Tutoring System for Precision Manufacturing
Curriculum Vitae
Updated August 2026Advisor: Prof. Stacy Marsella
Advisor: Prof. Stacy Marsella
Advisor: Prof. Mohammad Akbari
Evaluating large language models for co-speech gesture generation, and building an end-to-end LLM-driven gesture pipeline integrated into a virtual agent framework. Annotated a speaker-specific hand-gesture dataset used for evaluation.
Integrated a virtual human framework into a maternal health clinical AI system and developed an LLM-driven emotional expression module for the agent. Co-presented the company pitch to Pear VC, covering the virtual human architecture, design choices, and underlying research.
Supervisor: Dr. Sharon Mozgai. Developed a nonverbal behavior generation pipeline deployed in a Unity virtual agent platform, and designed a human-subjects study of an AR agent for workplace training.
Designed and ran a Wizard-of-Oz evaluation with 40 participants of an AR intelligent tutoring system for complex psychomotor tasks, with mixed-effects and qualitative analysis.
Neural networks for textual sentiment analysis; research on therapeutic chatbots.
Rubric-based grading for ~50 students; weekly office hours supporting research projects.
Python, R, C#
OpenAI API, PyTorch, Hugging Face, Scikit-learn, Pandas, NumPy, Unity, Git
NLP, LLMs — prompt engineering, fine-tuning, RAG
Study design, statistical analysis, Qualtrics, IRB write-up