Skip to content
NR
05/Case Study

AI Health Platform

RAG Chatbot & Gamified Nudges — NJIT Capstone

RoleLead UX & AI Engineer
Timeline2025
StackGPT-4, LangChain, RAG
StatusShipped
Impact

88% usability task success rate with AI chatbot achieving 4.17/5 satisfaction across diverse demographics.

Overview

Context

This capstone project represents the intersection of AI engineering, product design, and health technology. I led both the UX research and AI implementation, building a platform that provides personalized health guidance through a RAG-powered chatbot while using gamified behavioral nudges to encourage sustained engagement.

Challenge

The problem

Health information needs to be accurate, sensitive to demographics, and accessible. The challenge was building an AI-powered health platform that could provide personalized, safe guidance while being genuinely usable across diverse populations — and proving this through rigorous usability testing, not assumptions.

Approach

How I built it

01

Conducted 15+ stakeholder interviews and created journey maps to understand diverse health information needs

02

Built an 80+ component Figma design system with WCAG 2.1 AA compliance

03

Implemented LangChain RAG chatbot with 15+ prompt iterations for safety, tone, and demographic alignment

04

Designed gamified AI nudges driven by behavioral data to encourage health engagement

05

Built full backend with 15+ database tables, 40+ REST endpoints, JWT authentication

06

Led 5-person Agile team through design, development, and testing cycles

Technical Decisions

Why these choices

15+ prompt iterations for safety

Health AI demands exceptional care in tone, accuracy, and safety. Each iteration addressed specific failure modes: demographic insensitivity, medical overreach, accessibility gaps.

Gamified nudges over passive content

Static health information doesn't drive behavior change. Gamified nudges based on behavioral data create sustained engagement without being manipulative.

Outcomes

What shipped

88% usability task success rate in formal testing
4.17/5 System Usability Scale score
WCAG 2.1 AA accessibility compliance
80+ component design system
40+ REST endpoints with JWT authentication
5-person Agile delivery successfully completed
Takeaways

What I learned

Health AI requires more prompt engineering iteration than any other domain — safety is non-negotiable
Design systems pay for themselves when building accessible, consistent UIs at scale
Usability testing reveals gaps that no amount of internal review catches