🏥 Dulub — Project Story

🌟 Inspiration

Dulub was inspired by a simple but critical problem: access to healthcare is still uneven, especially for people who face geographical, financial, or time constraints. Telemedicine solves part of this problem, but many platforms stop at video calls. I wanted to go further by integrating AI as an assistant, not a replacement, to help both patients and doctors work more efficiently.

Another key inspiration was accessibility. Navigating apps can be difficult for elderly users or busy clinicians, so I designed Dulub to support Voice AI commands, allowing patients and doctors to interact with the system hands-free—booking appointments, opening medical records, or joining consultations using voice.

📚 What I Learned

Building Dulub taught me how to design and implement a real-world, production-ready full-stack system. I learned how to:

• Architect a multi-role system with patients, doctors, and admins
• Implement JWT-based authentication and role-based access control
• Work with asynchronous backends using FastAPI and Motor
• Integrate AI services (text + vision) responsibly in healthcare
• Design scalable MongoDB schemas for complex relationships
• Balance UX, security, and performance in a healthcare context

I also learned that AI in healthcare must always be framed as supportive, not authoritative, reinforcing the importance of professional medical advice.

🛠️ How I Built the Project

Dulub was built using a modern, scalable tech stack:

• Frontend: React 18 + TypeScript + Material UI for a clean, responsive healthcare UI
• Backend: FastAPI with async MongoDB (Motor) for high-performance APIs
• Database: MongoDB Atlas for flexible document-based storage
• AI: Google Gemini 2.5 Pro for symptom checking, OCR, and medical report analysis
• Video: Jitsi Meet for secure, registration-free video consultations
• Voice AI: Voice-driven commands layered on top of core workflows
• Deployment: Docker + Kubernetes with Nginx Ingress

The system follows a clear workflow: [ \text{Authenticate} \rightarrow \text{Authorize} \rightarrow \text{Process Request} \rightarrow \text{Respond Securely} ]

Each appointment generates a unique video room, medical records are securely stored, and dashboards are dynamically rendered based on user roles.

⚠️ Challenges I Faced

One of the biggest challenges was designing healthcare-safe AI behavior—ensuring that AI responses are informative without being misleading. Another challenge was managing role-based logic cleanly across both frontend and backend without duplicating rules.

Handling file uploads, OCR processing, and performance optimization in an async environment was also complex. Additionally, integrating multiple external services (AI + video conferencing) while keeping the system stable required careful error handling and testing.

Despite these challenges, the project pushed me to think like a system designer, not just a developer.

🚀 Reflection

Dulub represents my growth in building end-to-end systems that combine technology, AI, and human-centered design. It’s not just a telemedicine app—it’s a step toward smarter, more accessible healthcare powered by responsible AI.

Built With

• fastapi
• gemini3
• mongodb
• python
• react
• supabase
• typescript