• 

XVoice

Inspiration

Healthcare is one of the few places where communication failures can have life-changing consequences. During medical emergencies, patients often struggle to advocate for themselves due to language barriers, disabilities, stress, pain, cognitive overload, or an inability to speak. We wanted to build a solution that ensures patients never lose their voice when they need it most.

We were inspired by a simple question:

What if every patient could carry their voice, medical history, and preferences with them wherever they receive care?

This led us to create XVoice, an AI-powered patient communication layer that helps bridge the gap between patients and healthcare providers.

What it does

XVoice allows patients to create a portable healthcare profile by uploading insurance cards, medical records, prescriptions, test reports, and other healthcare documents or by simply speaking naturally in their preferred language.

Using AI, XVoice automatically:

• Extracts and organizes important medical information
• Identifies medications, allergies, conditions, insurance details, and emergency contacts
• Creates a patient-owned health profile
• Makes healthcare information searchable and accessible

Patients can generate a QR code that providers can scan during a healthcare visit.

Providers gain access to:

• A patient summary dashboard
• Medical history and uploaded records
• Previous visit information
• Current patient concerns
• AI-assisted patient Q&A

XVoice also supports real-time multilingual communication. Providers and patients can communicate through AI-powered voice translation, and when a patient cannot speak, XVoice can act as a communication proxy using only verified patient information and uploaded records.

Every interaction is logged in a live visit timeline, creating a shared source of truth throughout the encounter.

How we built it

We built XVoice using a modern AI-first architecture:

Backend

• FastAPI
• Python
• SQLite
• UV package management

Frontend

• React
• TypeScript
• Vite

AI

• Grok API
• Grok Vision for document understanding
• Grok Voice for multilingual communication
• RAG-style retrieval over uploaded patient records

Core Features

• QR-based patient sharing
• AI-powered document classification
• Medical record indexing and retrieval
• Source-grounded patient agent
• Visit dashboard and event timeline
• Real-time communication workflows

When a document is uploaded, XVoice extracts text, classifies the document type, generates tags, extracts key healthcare facts, and indexes the content so it can later answer questions using only information found in the patient's records.

Challenges we ran into

One of the biggest challenges was balancing usefulness with safety.

Healthcare applications cannot simply rely on AI-generated answers. We needed to ensure that:

• Information remained source-grounded
• Unknown information stayed unknown
• The system never fabricated medical facts
• The AI never acted as a doctor
• The AI never provided consent on behalf of a patient

Another challenge was handling the wide variety of healthcare documents. Insurance cards, lab reports, discharge summaries, prescriptions, and visit notes all have very different formats and structures.

We also had to design workflows that could work without requiring integration into existing hospital systems, making the experience simple enough to use during a hackathon while still demonstrating real-world value.

Accomplishments that we're proud of

We are proud that XVoice goes beyond simple translation and becomes a true patient advocacy platform.

Some highlights include:

• Creating a patient-owned healthcare profile that can be shared anywhere via QR code
• Building an AI-powered system that automatically understands uploaded healthcare documents
• Supporting multilingual communication between patients and providers
• Designing a source-grounded patient proxy that only answers from verified information
• Creating a live visit dashboard that keeps both patients and providers aligned throughout a healthcare encounter

Most importantly, we built a product that addresses a real and often overlooked problem: patients losing their ability to communicate effectively during critical moments of care.

What we learned

We learned that communication challenges in healthcare are much broader than language translation.

Patients often struggle because:

• They cannot remember their medical history
• Their records are fragmented
• They are overwhelmed during emergencies
• They have disabilities or accessibility needs
• They are unfamiliar with the healthcare system

We also learned that trust is essential in healthcare AI. The most useful answer is often not a confident answer—it is an honest answer that clearly communicates what is known, what is unknown, and where the information came from.

What's next for XVoice

Our vision is to make XVoice a universal patient communication layer that works across the healthcare ecosystem.

Future plans include:

• Direct integration with electronic health record systems
• FHIR-based health data connectivity
• Advanced multilingual support
• Caregiver and family collaboration tools
• Accessibility-focused experiences for patients with speech and hearing impairments
• Mobile applications for patients and providers
• Real-time healthcare interpreter workflows
• Emergency access mode for urgent care and hospital settings

Ultimately, we believe every patient should be able to carry their voice, history, and healthcare context wherever they receive care. XVoice is our first step toward making that future possible.

Built With

• grok
• python