Breez Dental

‹

›

Secure & Compliant Medical Apps

Developing applications for healthcare requires more than technical skill. It demands a deep understanding of data security and regulatory compliance. Handling sensitive patient data is a core responsibility, and we architect our platforms with security as the first priority.

Our work on the Bosco Dental Studio AI platform is a prime example. We developed an innovative SaaS solution to automate clinical note-taking for dentists, processing highly sensitive patient information.

The system was engineered for high-level security from the ground up, ensuring all patient notes and audio recordings are safely processed and stored. Our development process strictly adheres to all relevant data protection guidelines, including GDC and GDPR. We deliver the cross-platform application as a fully EV-signed and secure MSI installer, suitable for clinical environments.

Smart AI Integration

The AI landscape is evolving rapidly. Relying on a single AI provider for a core business function is a significant risk, so we design our AI products to be future-proof with a modular, flexible architecture.

The server architecture was specifically designed to readily integrate, evaluate, and switch between different AI transcription and NLP services. This multi-model strategy is essential for both performance and cost-effectiveness:

• We integrate and benchmark multiple speech-to-text services (FalAI, AssemblyAI, Deepgram) to find the optimal balance of speed, accuracy, and cost for a clinical environment.
• We test and deploy various large language models for note generation (OpenAI, Claude, Gemini) while navigating use-case restrictions related to medical use.

This approach ensures our clients can consistently leverage the best-performing and most cost-effective technologies without re-engineering the core product.

From Uncertainty to Innovation

Many groundbreaking projects begin with significant technological uncertainties. The initial challenge here was clear: generic, off-the-shelf transcription tools were unsuitable for a noisy clinical setting with multiple speakers.

We identified two core uncertainties and resolved them systematically:

• It was uncertain if a dual-microphone system could reliably differentiate a dentist’s voice from a patient’s in a live clinic. We conducted a controlled trial with 3 practitioners and approximately 50 patients. By analysing this real-world data, we proved our dual-channel approach was viable and provided a significant improvement.
• It was uncertain if we could process audio in real-time using a custom algorithm without losing clinical accuracy. We used an iterative development process, testing algorithm iterations in the lab before deploying the best options in a two-week field test.

This R&D process resolved fundamental uncertainties and created a secure, high-performance service that exceeded the initial MVP, leading to further client funding and platform expansion.