Implementation of elementary learning and inference functions on a simple recommendation model. Selection of an LLM allowing interaction with the user. Implementation of the use of an LLM in ReAct mode, after fine-tuning of the LLM, using the available functions. Comparison of several possible approaches. Creation of a demonstrator on a recommendation application, and test on an interaction with a user. All our internships are approved and subject to a bonus, the amount of which is determined according to your level of studies.