We deployed a diploma verification process for a major CAC 40 group using Llama 2 on-premise and in just a few weeks under the constraints of multiple countries/languages and data.
This automation reduced regulatory control time from 4 days to 2 hours.
Over 1500 contracts processed each quarter
98% accuracy in verification
2 to 3 hours of human work time instead of 4 days before AI
A major player in the CAC 40, Eméis (Orpea) is a global leader in medical care and support with over 78,000 healthcare professionals. For more than 30 years, its international expertise has covered the entire care chain, from specialized medicine to personalized support
Verifying diplomas, contractual clauses, and visas represented a major challenge for Eméis' HR teams. The manual process required three to four days to process the files of 1,500 employees, tying up considerable resources and slowing administrative efficiency.
The complexity was amplified by the multilingual nature of documents, written in French, English, and German. The varying quality of documents, whether scanned or photographed, added an extra layer of difficulty to processing. Faced with these challenges, the need for an automated solution capable of ensuring GDPR compliance and sensitive data security became evident.
Our approach centered around an integrated solution, deployed directly on Eméis' infrastructure. The platform combines advanced OCR technology for document digitization with generative AI Mistral-Nemo for information analysis and validation.
The system intelligently processes documents in different languages, adapts to varying scan qualities, and accurately extracts relevant information. The entire solution is deployed on-premise, ensuring maximum security and full compliance with GDPR requirements.
Advanced multilingual OCR
Processing documents in multiple languages
Adaptation to different scan qualities
Reliable extraction of key information
Generative AI with Mistral-Nemo
Intelligent content analysis
Automatic information verification
Validation according to predefined rules
Secure deployment
Installation on Eméis' internal servers
GDPR compliance guaranteed
Maximum protection of sensitive data
The project was conducted over four months by a tight team consisting of an AI engineer and a product manager. This agile configuration allowed for rapid iterations and continuous adaptation to feedback. Development was done in close collaboration with Eméis teams, allowing the solution to be refined under real conditions.
The adopted methodology favors short development-test-improvement cycles, with particular attention paid to the specificities of the medical sector and data security requirements.
Key phases
Analysis of needs and constraints
Platform development
AI model integration
Testing on real data
On-premise deployment
Continuous optimization
The impact of the solution on Eméis' operations is significant.