The context
Supporting patient triage and referral in screening centres while structuring medical information for the teams. AI is an assistance tool, not a replacement for dermatologists.
My role
I lead the full cycle: architecture, development, deployment and technical reviews. I recruit and coordinate external developers, and work with dermatologists and the LIS laboratory to connect clinical needs with technical decisions. I also provide industrial supervision for a CIFRE PhD with LIS.
The technical approach
The triage model combines images and clinical metadata. OCR + LLM pipelines structure medical reports into database records. Internal tools, APIs and LLM agents support operations; the infrastructure includes AWS, Vercel and Docker.
Deployment and use
The triage model is deployed in 6 centres. Since April 2026, it has been used to analyse over 100,000 lesions involving more than 10,000 patients. These volumes illustrate its operational use in dermatological triage workflows.
Governance and next steps
I contribute to health data governance and the preparation of documentation towards CE marking as a class IIa medical device. Code, patient data and detailed architectures remain confidential.