Projects

TriageAI is an AI-powered patient triage agent delivered through a web-based chat interface. A patient lands on the chat interface and describes what's going on in their own words. The AI agent asks clarifying follow-up questions, gradually building a structured clinical picture from an unstructured conversation. At the end of the interaction, the agent produces a written summary for the provider and a recommendation on both the urgency of the case and which department the patient should be routed to. Under the hood, the agent is not a static decision tree: it dynamically decides what information is still missing based on a clinical intake schema the team defines, and it can pull in outside clinical reference material (e.g., openFDA, clinical guidelines, or a local knowledge base) to decide which follow-up questions matter for a given complaint.

Session Prep Pro is a full-stack clinical productivity tool built for licensed therapists and counselors in private practice. Clinicians maintain a private roster of clients (using anonymized identifiers such as initials or pseudonyms), each with clinician-authored background context. Before a session, the clinician pastes in raw notes — progress notes, EHR exports, prior session summaries — and the system uses a large language model (GPT-4o via the OpenAI API) to generate a structured Pre-Session Brief with four sections: Sensitivities — clinically sensitive topics or dynamics extracted from the notes, each tagged with an urgency level (high / medium / low) Key Themes — recurring therapeutic themes across session history Follow-ups — specific items the client raised that need to be checked on next session Opening Recommendations — concrete, actionable language for how to open the upcoming session Brief generation runs asynchronously: the server returns immediately, processes the AI call in the background, and the frontend polls every 2 seconds until the brief is complete. A dashboard gives the clinician an at-a-glance view of client count, briefs generated, and recent activity. This is a clinician-side workspace only — it is not a therapy tool or patient-facing product. It is purely internal, scoped to each individual therapist's own clients, and designed entirely around the pre-session preparation workflow. A working prototype already exists (built on Replit.ai) demonstrating the core flow end-to-end, including a monorepo architecture (pnpm workspaces), contract-first API design (OpenAPI + codegen), Clerk-based multi-tenant authentication, and a React/TypeScript frontend. This gives the capstone team a running start, with the bulk of the work centered on hardening, extending, and validating the system for a real pilot.

Between Sessions is a secure, passcode-gated AI companion web application designed for mental health clients to use between scheduled therapy appointments. The application provides a structured, supportive conversational interface powered by a large language model, allowing users to process thoughts, track their emotional state, and engage in reflective journaling without direct clinician involvement.

Practice Hub is a full-stack web application that serves as a centralized credentialing profile hub for licensed therapists in private practice. The clinician builds and maintains a single, structured master profile containing every piece of information that credentialing bodies and insurance panels repeatedly ask for: NPI number, practice addresses, active state licenses with expiration dates, professional liability insurance details, educational degrees, contracted insurance panels, and a professional vitae. From this master profile, the application generates platform-specific templates — formatted, copy-paste-ready documents tailored to major platforms like Headway, Alma, and TherapyNotes/Unite Native, as well as smaller group practices and EHR-based organizations. Additionally, Practice Hub exports a standardized, machine-readable data file (JSON and/or PDF) that can be handed directly to a credentialing coordinator at a new organization, eliminating the back-and-forth of credentialing packets. The application is built on a React and Vite frontend, an Express 5 REST API backend, PostgreSQL with Drizzle ORM, and Clerk for per-therapist authentication and data isolation, with the full API defined in an OpenAPI 3.1 specification from which TypeScript client hooks are auto-generated.

Universities collect extensive enrollment, academic, demographic, and engagement data, but advisors and administrators often lack timely, integrated insights into students who may be at academic risk. In collaboration with the Oregon State University College of Engineering, students will design and develop a student success intelligence platform using historical institutional data. The platform will support prediction of student outcomes, analysis of risk factors, and data-driven decision-making. Students will work through an end-to-end workflow including data engineering, exploratory analysis, predictive modeling, evaluation, visualization, and deployment. Depending on team interests, they may also explore large language models, explainable AI, agentic AI, and interactive decision-support dashboards.