AI Patient Education Chatbot for Ophthalmology
~75%
2–4 months
Not disclosed
Medical staff spent excessive time on repetitive patient education about procedures like cataract surgery. Information needs varied across practices, and any production deployment had to meet HIPAA requirements.
PressW built Sophia, an AI conversational chatbot tailored to ophthalmology practices, with a NextJS front end, OpenAI Whisper for speech-to-text, Eleven Labs for speech synthesis, and ChatGPT fine-tuned to each practice's tone and procedures.
Working with the practice, PressW built a secure cloud architecture on Google Cloud, Firebase, and Pinecone for knowledge management, prioritizing HIPAA compliance and data separation across practices.
Medical practices that want to automate repetitive patient education while keeping interactions personalized, compliant, and practice-specific.
- Google Cloud (hosting environment)
- Firebase (application data layer)
- Pinecone (vector store for practice knowledge management)
- Per-practice data separation designed for HIPAA compliance
- NextJS front end serving the patient interface
- Patient voice input to OpenAI Whisper speech-to-text to the chatbot
- Chatbot response to Eleven Labs speech synthesis to voice output
- Practice-specific procedure and tone content in a Pinecone knowledge store, retrieved at answer time
- Per-practice data separation enforced across the shared cloud architecture
400+ Hours/Month Saved (Projected)
HIPAA-Conscious, Multi-Practice
Text and Voice Patient Education
Bryson Greenwood
The practice deployed a conversational AI chatbot fine-tuned to its own procedures and tone, handling repetitive explanations of procedures such as cataract surgery in both text and voice. Staff time spent on patient education is projected to fall by more than 400 hours a month at a single practice.
ChatGPT fine-tuned to each practice, OpenAI Whisper for speech-to-text and Eleven Labs for speech synthesis, with a NextJS front end running on Google Cloud, Firebase and Pinecone.
An estimated 400+ hours of combined patient-education time saved per month in a single practice. Patients can interact by text or voice, and the secure cloud architecture keeps data separated across practices for compliant deployment.
The record does not state a timeline. The work covered building a secure cloud architecture on Google Cloud, Firebase and Pinecone with HIPAA compliance and data separation as the priority, then fine-tuning the chatbot per practice.
Medical practices that want to automate repetitive patient education while keeping interactions personalized, compliant, and practice-specific.
Have a similar challenge?
