The Context

A team of dental professionals set out to transform the clinical experience after growing increasingly frustrated with outdated software and the inefficiencies of manual patient file transcription. Their vision: a connected dental mirror that significantly reduces the need for manual data entry, allowing practitioners to focus more on patient care and deliver higher-quality outcomes.

Following several years of hardware-focused research and development, they engaged us to design and develop a modern medical records platform. We led the complete technology initiative from discovery through to development—leveraging natural language processing (NLP), computer vision, firmware integration, robust security practices, and more. Here’s how the solution came to life.

The Challenges

The device features a compact design with small, highly efficient components, optimized for both performance and minimal space usage. It includes a built-in camera and light source that operate within a patient’s mouth, streaming a live feed to a connected tablet. Due to the compact hardware constraints, our development team had to ensure extremely precise programming to capture and transmit the video feed with minimal latency or lag.

The AI-powered platform was designed to incorporate advanced computer vision capabilities, enabling the real-time processing of images and videos for machine learning purposes. Key automated functionalities included:

• Detecting when the device is in the patient’s mouth and adjusting the LED lights accordingly

• Rotating and aligning images

• Enhancing image clarity and quality

Additionally, natural language processing (NLP) was integrated to support real-time speech-to-text conversion. This allowed for automatic transcription of spoken notes—such as tooth chart readings—directly into the patient’s digital records, significantly reducing the burden of manual documentation for dental staff and allowing them to focus more on patient care.

Given the platform’s use of sensitive healthcare data in both Canadian and U.S. markets, full compliance with HIPAA and GDPR standards was critical. Specific data storage policies were implemented to meet regulatory requirements, including mandates that patient data from Canada be stored within Canadian borders. End-to-end encryption was applied to all data transmissions and storage, ensuring maximum security and privacy—so much so that even the development team cannot access any patient information.

Task

The goal was to create a solution that felt simple yet complete. Designing for a tablet environment allowed us to focus on essential features, delivering a streamlined and intuitive user experience. During a six-week Discovery phase, we iterated through several proof-of-concept (POC) stages to validate core functionality before moving into full-scale development—covering everything from mobile integration and secure data hosting to 3D modeling and hardware implementation.

The hardware component of the solution uses gRPC to enable fast and efficient API-based communication, allowing seamless integration with external services and applications. A significant technical challenge emerged when we needed to control the hardware directly from a web application. The hardware operates using Wi-Fi Direct, which is not accessible through traditional web browsers due to its requirement for low-level system access on the tablet.

To address this, we engineered a lightweight custom micro-backend service that operates as a local web server. This intermediary layer enables hardware control via the web app, including video streaming, LED management, and other device functionalities—bridging the gap between hardware and browser limitations.

For data storage, we selected Amazon DynamoDB, a serverless, fully managed NoSQL database known for high scalability and single-digit millisecond response times—perfect for performance-sensitive applications. Its pay-per-use pricing model eliminated unnecessary overhead, making it a cost-efficient choice. We also took advantage of DynamoDB’s rapid data processing capabilities for machine learning use cases, while its built-in encryption supported the high standards required for HIPAA compliance. DynamoDB’s performance and security were instrumental in delivering a responsive and regulation-ready platform.

The project is built on a serverless architecture powered by AWS Lambda, allowing the application to dynamically scale with demand. This architecture ensures zero infrastructure cost during idle periods, as resources automatically scale down when not in use. The result is a highly efficient system where costs are only incurred when resources are actively being consumed.

Technology Stack

Frontend:

• Platform: Android (Tablet-based application)

• Framework: Kotlin / Jetpack Compose or Flutter (for cross-platform UI)

• UI Libraries: Material Design, Lottie for animations

Backend:

• Language: Node.js

• Framework: Express.js

• Architecture: Microservices-based architecture

• Communication Protocol: gRPC for low-latency hardware communication

AI/ML:

• NLP Engine: Google Cloud Speech-to-Text 

• Computer Vision: TensorFlow Lite (for image rotation, enhancement, and mirror detection)

• ML Deployment: TensorFlow Lite for edge deployment on tablets

Database:

• Primary Database: Amazon DynamoDB (serverless NoSQL for scalability)

• Encryption: AES-256 encryption at rest and in transit

Cloud Infrastructure:

• Platform: AWS

• Services:
  AWS Lambda (for serverless compute)
  Amazon S3 (secure storage of images/videos)
  API Gateway (for routing client requests)
  AWS IAM (for fine-grained access control)

Security & Compliance:

• Data Compliance: HIPAA, GDPR

• Authentication: OAuth 2.0

• Data Localization: AWS Regions configured per jurisdiction (e.g., Canada Central for Canadian data)

Firmware & Hardware Integration:

• Firmware Communication: WiFi Direct

• Micro-backend Service: Embedded local web server to bridge hardware commands from the tablet

• Protocols: WebSockets / HTTP for real-time control (LEDs, camera feed)

Monitoring & Analytics:

• Tools: CloudWatch, Sentry (for error tracking and logging)

• Dashboards: Custom admin dashboard for monitoring usage, hardware diagnostics

Result

After a year of focused development, Version 1 of the medical records platform has been successfully completed.

Dental practices adopting the solution will receive a fully integrated tech package, including Wi-Fi-enabled hardware with interchangeable mirror heads for easy sterilization, a tablet preloaded with the custom software, and a mounting arm for convenient in-practice use. The platform’s photo, audio, and video recording capabilities empower dental professionals to visually explain procedures, educate patients on oral health maintenance, and easily compare current findings with previous records for more informed care.