■ Project Overview
An AI-Powered Intelligence and Patient Personalization Platform for Aesthetic Clinics
8 members × 6 months
Project Manager (PM) / Business Analyst (BA) / Developers / Quality Control (QC)
Requirements Definition / UI & UX Design / Development / Testing / Deployment / Maintenance & Operations
■ Technology Stack
AI & Image Analysis: Vision Transformer (ViT) / MediaPipe / OpenCV / GPT-4o Vision
LLM & RAG: GPT-4o / LangChain / Qdrant
Backend: Python (FastAPI) / PostgreSQL / Redis
External Integration: LINE Messaging API
Frontend: Flutter / React / TypeScript
Data Platform: BigQuery / Airflow / Azure Blob Storage
Cloud & MLOps: Azure (AKS / Azure OpenAI) / MLflow / GitHub Actions
■ System Features
Integrated AI platform for aesthetic medicine, combining image analysis, LLMs, and CRM capabilities.
RAG-based treatment recommendations with automated customer engagement via LINE integration.
Decision-support AI design compliant with medical regulations, ensuring a clear distinction between diagnosis and assistance.
Standardization of facial images (angle and lighting correction).
Facial landmark extraction using MediaPipe.
Vision Transformer (ViT) analyzing 12 skin-related indicators.
GPT-4o Vision generating comprehensive reports in Japanese.
87% concordance rate with physician evaluations.
Integration of customer data, skin analysis, and medical history.
Retrieval of optimal treatment protocols using Qdrant.
GPT-4o generates three evidence-based treatment options with explanations.
Utilized as clinical decision support, compliant with medical regulations.
Time-series comparison using facial alignment techniques.
Quantification of improvement rates.
Visualization through the application to enhance patient experience.
Scenario-based messaging after treatment.
Immediate post-treatment care guidance, progress monitoring, and re-visit promotion.
■ Project Outcomes
Skin analysis accuracy: 87% agreement with physician assessments.
Reduced consultation time: from 45–60 minutes to 20–25 minutes.
Expansion decision: After a PoC at two clinics, deployment across all branches was approved within 8 weeks.
Patient satisfaction increased from 4.1 to 4.8 out of 5.
Enhanced treatment understanding through clear visualizations.
40% reduction in physicians’ explanation time.
Easier identification of problem areas, improving clinical efficiency.
Revisit rate increased from 42% to 56% (+34%).
Customer Lifetime Value (LTV) increased, with revenue per patient rising by 29%.
Conversion Rate (CVR) improved from 38% to 51%.
Highly praised for compliance with medical regulations, particularly the clear distinction between diagnosis and decision support.
Recognized for UX quality tailored to the Japanese market.
Approved for expansion to the next development phase.
