Wearables + LLMs
RELIEF Study
Designed the RELIEF Study, integrating wearable AI and LLMs for real-time personalized stress interventions in the field.
Selected Projects
A curated selection of research systems, modeling efforts, and applied machine learning projects spanning wearable sensing, mental health, personal informatics, survival modeling, NLP, and interactive AI.
Field Study
MOODS Study
Led the nationwide MOODS Study with 122 participants, achieving a 10% reduction in stress and 14 behavior modifications (CHI 2024).
Personal Informatics
Personal Informatics System
Developed 16 novel visualizations to promote self-reflection and engagement.
Survival Modeling
Stressor Longevity for Just-in-Time Interventions
Modeled real-life stressor durations from 100-day wearable data using Weibull survival analysis models; estimated half-lives ranged from 30–290 minutes.
Behavior Modeling
Behavioral Change Prediction
Designed and developed survival analysis models using Random Forest and Gradient Boosting with a C-index of 0.78 to predict user-initiated behavioral changes for personalized stress management.
Prompting Systems
AI-Triggered Prompting (ATP)
Developed a framework to optimize prompt timing and predict personalized stressor listings, minimizing interruptions in real-world settings.
Foundation Models
Pulse-PPG
Co-author of Pulse-PPG, the first open-source PPG foundation model trained on raw PPG data from a 100-day study (Paper).
Evaluation Dashboard
LLM-Assisted Model Ranking
Designed a dashboard ranking models (GPT-4o, Claude-Sonnet, LLaMA3-8B) for stress intervention generation (GitHub).
Low-Resource NLP
NER for Nepali Sentences
Developed a framework for evaluating Named Entity Recognition (NER) tasks in Nepali (GitHub).
Clinical Collaboration
Sickle Cell Pain Study
Analyzed opioid use in sickle cell patients with Johns Hopkins University, exploring affect and sleep impact on pain (Journal of Pain 2024).
Recommendation Systems
Movie Recommendation System
Built a recommendation system using content-based and collaborative filtering (GitHub).
Social Media NLP
COVID-19 Vaccine Sentiment Analysis
Developed a sentiment analysis model analyzing emotions in COVID-19 vaccine tweets (GitHub).
Information Retrieval
Custom NLP Search Engine
Built a search engine using NLP for University of Memphis staff and students (GitHub).