Inspiration
The idea for MediScan AI was born from a very personal experience: my co-founder cut his thumb while cooking and had to rush to the ER. In the chaos and urgency of the emergency, he later realized he’d been overcharged but had no way to contest the bill at the time—he simply didn’t know what to do. When he began researching, he discovered that this wasn’t an isolated incident; millions face similar situations every year, with little recourse and even less transparency. The scale and impact of medical billing errors and overcharges were too big to ignore, and we knew technology could help.
What it does
MediScan AI is an on-device app that empowers patients to scan, review, and understand their medical bills and insurance documents instantly. Using edge-based OCR and AI, it extracts codes and charges, flags potential errors or overcharges, explains complex medical terms in plain language, and helps users generate dispute letters—all while keeping sensitive data private and secure on their device.
How we built it
We built MediScan AI using a React-based frontend, integrating on-device OCR (Tesseract/EasyOCR) and lightweight NLP for medical code extraction and explanation. The backend leverages Bolt’s GenAI SDK for rapid prototyping, with local storage and AES-256 encryption to ensure HIPAA compliance. We used Bolt’s LLM capabilities for generating patient-friendly explanations and dispute templates. All processing happens on-device, minimizing privacy risks and ensuring no PHI leaves the user’s control. We iterated quickly, using Bolt’s visual editor and prompt-driven development to refine features and UI based on real user feedback and hackathon guidelines.
Challenges we ran into
HIPAA compliance: Ensuring all PHI stayed on-device required careful architecture and encryption.
OCR accuracy: Medical bills come in many formats; tuning the OCR pipeline for high accuracy was tough.
Medical code complexity: Mapping CPT/ICD codes to plain language and detecting billing anomalies required building and maintaining large, up-to-date dictionaries.
User education: Making complex medical and insurance information understandable to non-experts was a constant UX challenge.
Time constraints: As a two-person team with limited time, we had to ruthlessly prioritize features for the MVP
Accomplishments that we're proud of
Delivering a working MVP that can scan and analyze real medical bills on-device, flag errors, and generate helpful explanations and dispute letters.
Building a secure, HIPAA-conscious workflow with all sensitive data processed and stored locally.
Creating a clean, user-friendly interface that demystifies medical billing for everyday users.
Rapidly iterating using Bolt’s platform to deliver a polished demo within the hackathon timeline.
What we learned
The scale of medical billing errors and their impact on patients is staggering—and most people have no tools to help.
Edge AI and on-device processing are not just technical advantages but essential for privacy and trust in healthcare.
User education and clear explanations are as important as technical accuracy in building trust and usability.
Prompt-driven and LLM-assisted development (via Bolt) can dramatically accelerate building complex, real-world solutions.
What's next for Mediscan.ai
Expand error detection to cover more billing scenarios and support additional document types (EOBs, prescriptions).
Integrate more advanced, quantized LLMs for richer on-device conversations and insurance Q&A.
Build partnerships with advocacy groups, clinics, and employers to reach more users.
Launch a freemium model, with advanced features for premium users and potential B2B integrations.
Continue refining the UI/UX based on user feedback and evolving healthcare regulations.
Ultimately, empower millions to take control of their healthcare finances with clarity and confidence.
Log in or sign up for Devpost to join the conversation.