Portfolio

This project uses Natural Language Processing (NLP) and Large Language Models (LLMs) to collect and extract mental health scientific literature from multiple databases, enabling the construction of a comprehensive knowledge graph for research and analysis.

This project uses digital storyboards to co-design with patients with kidney disease, their caregivers, and clinicians in building a digital health application. The goal is to support all stakeholders in shared decision-making during the healthcare process.

Developed and evaluated a multi-agent conversational recommender system, where two distinct AI film critics engage in dynamic banter to help users choose movies. Built using multiple language model instances, the system employs voice interaction (via speech-to-text integration), persona-driven prompt engineering, and interaction logic that reinjects critic personas and encourages contrasting responses. Inspired by real-world commentator dynamics, the system maintains critic individuality through prompt rotation, tonal cues, and reactive dialogue structuring. Results show that these methods significantly enhanced user engagement, realism, and recommendation experience.