Zibi

AI powered app that helps med students ace the USMLE with smart flashcards, voice quizzes, and personalized learning, anytime, anywhere.

Comment

Zibi – Project Story

Inspiration

While my wife was preparing for the USMLE boards, I saw firsthand the intense pressure, mental fatigue, and inefficiencies in traditional study methods like flashcards and longform review. As a software engineer, I knew technology especially AI and voice could play a role in making this process smarter, faster, and more supportive for medical students.

What It Does

Zibi is an AI-powered study companion designed specifically for USMLE prep. It helps students:

  • Generate personalized flashcards from notes, lectures, or uploaded content
  • Review with a built-in spaced repetition system
  • Study hands-free with voice quizzes powered by Amazon Polly and Transcribe
  • Get instant explanations from an AI tutor
  • Track progress, performance, and retention patterns

How We Built It

  • Frontend: Remix framework for fast, server-rendered web experience
  • Backend: FastAPI (Python) deployed via AWS Fargate
  • AI: Amazon Bedrock to generate and explain flashcards
  • Voice: Amazon Polly and Transcribe for spoken quizzes and interaction
  • Auth: Amazon Cognito
  • Database: Aurora Serverless (PostgreSQL)
  • Storage: S3 for user files and audio content
  • APIs: RESTful endpoints for flashcards, reviews, voice sessions, and progress

Challenges We Ran Into

  • Designing a voice-based interaction that felt natural and smooth
  • Balancing accuracy and simplicity in flashcard generation from unstructured text
  • Integrating multiple AWS services securely and efficiently
  • Creating an interface that is functional under time constraints, yet intuitive

Accomplishments That We're Proud Of

  • Built a full-stack architecture that is scalable, modular, and voice-enabled
  • Seamlessly integrated AI generated content and spaced repetition
  • Designed a workflow where users can learn visually, verbally, and contextually
  • Personalized the experience for real-world USMLE prep scenarios

What We Learned

  • Medical education is underserved in the area of practical, AI-enhanced tools
  • Voice-first learning opens new doors for engagement and accessibility
  • Remix paired with serverless deployment (Lambda) is powerful for dynamic apps
  • User feedback from a real med student helped us iterate meaningfully

What's Next for Zibi

  • Expand flashcard source support (PDFs, audio notes, etc.)
  • Add mobile app using React Native for even greater accessibility
  • Release a premium tier with expert-reviewed flashcard packs
  • Integrate peer-to-peer decks and study groups
  • Launch to a broader medical student audience ahead of the next USMLE cycle

Built With

Share this project:

Updates