Project Community Pathway Intelligence Platform - Turning Interview into Structured Insights Program DUIRI - Discovery Undergraduate Interdisciplinary Research Internship Term Summer 2026 Status Accepted Research Area Global Health Description Substance misuse—including alcohol, tobacco, and other drugs—creates substantial health and societal harms worldwide. Yet many communities continue to struggle to prevent and respond to substance use, often citing limited healthcare access, workforce shortages, transportation barriers, and fragmented service delivery as persistent barriers. Building a recovery-oriented system of care (ROSC) requires a practical, system-level understanding of how people actually move through local supports over time, including where they disengage, what prompts re-engagement, and which cross-system handoffs create gaps. Community organizations and coalitions frequently seek this insight by conducting interviews with people who have lived experience of substance use and then analyzing their recovery pathways. However, a major limitation of this interview-based approach is the difficulty of converting rich, narrative data into a clear, holistic, and actionable picture of the local system. Interview narratives are often story-based, fragmented, and non-linear; they rarely follow a simple “use ? treatment ? recovery” progression and instead describe repeated relapse–recovery cycles, interrupted treatment episodes, periods of informal recovery outside formal services, and parallel involvement with healthcare, justice, and social service systems. When many unique trajectories are combined using conventional summaries or simple flow diagrams, the result can become either oversimplified or an unreadable “spaghetti map,” obscuring common patterns and system bottlenecks that coalitions need to prioritize interventions. Therefore, there is a critical need for an advanced pathway mapping tool that translates qualitative interview narratives into standardized, coalition-ready pathway representations—highlighting recurring transitions, shared gaps, and cross-sector touchpoints—to support planning and implementation of community-based, evidence-informed actions. The full scope of this project contains multiple tasks: Task 1: Design an LLM-augmented interview-to-pathway translation engine that processes multi-modal unstructured data and converts them into structured objects and relations. Task 2: Build an interactive visualization platform for visualizing, exploring, annotating, and refining pathway maps with aid of an embedded graph analytics module. Task 3: Integrate AI frameworks for tailoring and quantification of community-specific pathway maps and a reinforcement learning framework for human-AI shared action-planning. During summer 2026, we will work on task 1, turning unstructured multimodal interview data into structed data for future processing and visualization. Supervisor Yingjie Chen Mentor Nan Kong Type of work required Software system development involving Python and AI integration. Websites Qualifications Strong Python programming skills and experience developing and integrating APIs with AI tools and services. GPA 3.0 or above. Credit Hours 3 Weekly Hours 40 (estimated)