DEV Community

Cover image for Graphing the Future: How We Built HealthCurve Nova AI for HackHazards 2026 🚀
AMAN RAJ
AMAN RAJ

Posted on

Graphing the Future: How We Built HealthCurve Nova AI for HackHazards 2026 🚀

#ai #devchallenge #showdev #namespace

Graphing the Future: How We Built Health Curve Nova AI for HackHazards 2026 🚀

Balancing my BCA coursework while diving into a massive buildathon is always a rush. HACKHAZARDS '26 runs as a two-month event from May to June 2026, giving us the time and space to ship a production-grade project instead of just a rushed prototype. With an incredible line-up of sponsors and a global pool of builders, my team and I set out to build Health Curve Nova AI.

Here is a look into our development journey, the challenges we tackled, and the tech that made it possible.

🛑 The Problem We Solved

Managing multiple prescriptions, understanding dosages, and maintaining a strict medication schedule is a daily struggle for millions, particularly the elderly. Missed or incorrect doses can lead to severe complications.

Health Curve Nova AI solves this through a dynamic OCR-based medication scheduling system. By simply snapping a picture of a medical prescription, our application uses AI to read the text, extract the medication details, and automatically generate a dynamic, conflict-free schedule for the patient.

🛠️ Our Tech Stack

To build a scalable and responsive application, we utilized a mix of powerful frameworks and tools:

  • Frontend: Expo (JavaScript/React Native) – Allowed us to rapidly develop a seamless, cross-platform mobile user interface.
  • Backend: Python (Flask) – We needed a lightweight, highly performant backend to handle our core OCR logic, AI agent routing, and API endpoints.
  • Database: Neo4j – The backbone of our data architecture, allowing us to map complex relationships between patients, medications, and time intervals.
  • Deployment & DevOps: Namespace and Docker – For containerization and streamlined deployment.

🤝 Partner Tracks Used

We specifically targeted the partner tracks provided by the hackathon sponsors to elevate our project. The hackathon is structured around core technology pillars, and we made sure to utilize them effectively:

  • The Neo4j Track: Traditional relational databases wouldn't cut it for our highly connected data. We leveraged Neo4j's graph database capabilities to map complex medication schedules and check for potential drug interaction nodes.
  • The Expo Track: We utilized the Expo framework to handle our frontend architecture, significantly cutting down our development time while ensuring a high-end UI/UX.
  • The Base44 & Namespace Track: We utilized Base44 to rapidly prototype our MVP and relied on Namespace for robust, scalable deployment.

⚙️ Implementation & Architecture

Our architecture follows a straightforward but powerful flow:

  1. The user takes a photo via the Expo frontend, which sends the image to our Flask backend.
  2. The backend triggers our AI/OCR pipeline to parse the messy prescription text into structured JSON data.
  3. From there, the backend constructs Cypher queries to interact with the Neo4j database. It checks for existing prescriptions, maps the new schedule, and flags any overlapping contraindications.
  4. Finally, the processed schedule is pushed back to the frontend, which sets up local push notifications for the user.

🧗 Challenges Faced

No hackathon is complete without a few late-night debugging marathons.

  • Taming the AI: Getting the OCR model to accurately read varying handwriting styles from doctors required extensive prompting and fallback logic.
  • Connecting the Graph: Wrapping our heads around Cypher queries and optimizing them for our Flask backend took some trial and error. My teammate, Aditya Raj, and I spent hours debugging our graph schemas to ensure the edge relationships between different drugs were mapped correctly.
  • State Management: Managing complex state across our Expo application while ensuring the UI remained snappy required several refactors.

🎉 Conclusion

Building Health Curve Nova AI for HackHazards 2026 pushed us to explore new architectural patterns, heavily utilize AI agents, and leverage graph databases in ways we hadn't before. It was a massive step up from a standard 24-hour sprint.

A massive thank you to the sponsors for the tools and the organizers for the platform! If you want a deeper behind-the-scenes look at the code and our development process, be sure to check out the devlog series on my YouTube channel, AmanCodeLogs.

Let me know your thoughts on our approach in the comments below! 👇

Top comments (0)