Context
Nos Quartiers ont du Talent (NQT), an association, supports graduate students in their professional integration through a mentoring scheme. Outside the working hours of mission managers, graduate students do not always have immediate answers to their questions.
The project consists in creating a chatbot taking on the role of mission manager assistant for the association, in order to respond in real time to these requests.
Our Intervention
The chatbot is added as a third participant in conversations on NQT messaging, which allows the mission manager to supervise the exchanges and complete them if necessary.
In order to be relevant in its responses, the chatbot is fed by the internal documentation of the association describing the mentoring process, by communication guidelines as well as by a database of reference exchanges.
The solution is based on a Java / Spring Boot / REST architecture, combined with a generative AI approach combining the Mistral LLM, a RAG logic and embeddings.
Results
A preliminary version of the chatbot, presenting positive results, is deployed and serves as support for:
- Demonstrations to the various internal stakeholders involved.
- Integration into NQT messaging chatrooms.
Optimization work has been carried out to reduce the energy impact and operating costs of the solution. A fine-tuning phase is planned to maximize the quality of responses once the service is put into production.
In the long term, the technical solution implemented will be generalized across the organization, with the migration of existing AI projects to this platform and the integration of new AI use cases.