MEDICAL AI

Custom-designed AI interface for a private healthcare center based on Meta LLaMA – 30% decrease in average consultation time.

02

Project & My Role

I led the design of a secure, in-house AI assistant for a private healthcare center. The goal was to replace external AI tools like ChatGPT with an internal, compliant system. I was fully responsible for UX strategy, interface design, and aligning the product with doctors’ daily routines inside the Apple ecosystem.

02

The Problem I Solved

Doctors were overwhelmed with manual tasks during patient visits: switching between apps, searching for data, calling colleagues, writing summaries from scratch. These interruptions extended consultation times and took focus away from patients.

To speed up their work, many staff members started using external AI tools like ChatGPT—copying sensitive data into public systems, which created serious GDPR and compliance risks.

The clinic needed a secure, in-house solution that would:

  • centralize medical, administrative and scheduling tasks

  • similar with their daily Apple-based apps

  • respect strict data privacy standards

  • and be so intuitive that it required no training.

02

Results I Delivered

⏱️ Reduced consultation time by ~30%

🧠 Enabled multitasking across patient care, admin, and scheduling

💬 80% of users rated the assistant as "essential" within two weeks

💻 Created a scalable interface used across medical, admin, and personal tasks

02

What I Designed & Why

I designed a modular, multitasking interface with separate AI „projects” for patients, resources, scheduling, and admin work. Each assistant had a unique prompt and knowledge base. I introduced:

  • Auto-created patient threads from EHRs for immediate access

  • Real-time voice recording with live transcription for visit summaries

  • File upload and contextual chat for reviewing documents or images

  • @mention logic to check colleagues’ availability and hand off patients

  • Admin dashboard for managing users, prompts, and content updates
    Everything was built to feel like native to macOS – clean, light, fast.

02

SUMMARY

This wasn’t a generic AI wrapper. I built a tool that fit doctors’ habits, respected their environment, and earned their trust. By combining strategic research, modular UX thinking, and native-level design, I helped turn a technical requirement into a daily-used assistant with real impact.

02

TAKEAWAYS

Every screen had to serve an actual use case – no fluff.

Separate AI brains made the tool clear and adaptable.

Doctors loved the product because it looked and behaved like an Apple app.

see more of my work

Document generation time reduced by 87.5%.

Saved about 10,000 € in dev time.

Daily quoting efficiency up by 190%.