Jonas Mouret

Complete development of a custom web application allowing sports coaches to create, schedule, and manage their group classes.

Clients can view the calendar, register online, and pay for their sessions through a secure payment system integrated with Stripe.

The application includes:

An intuitive administrator space for coaches: course creation, reservation management, and payment tracking.

A clear and responsive user dashboard for clients: registrations, history, notifications.

A modern architecture based on Django (backend) and React (frontend), with REST API integration and PostgreSQL database.

This project highlights my skills in full-stack development, payment solution integration, and ergonomic interface design.

In collaboration with the R&D team, I designed and developed an enterprise chatbot application based on a RAG (Retrieval-Augmented Generation) architecture.

This system aims to enhance the accuracy and reliability of responses generated by a language model (Mistral LLM) by relying exclusively on verified internal data.

The RAG pipeline I implemented combines:

A vector semantic search phase (via Qdrant) to identify relevant documents;

A contextual generation phase orchestrated by Mistral AI, restricted to information actually present in the knowledge base.

Data comes from specialized professional databases in architecture, pre-processed, chunked, and enriched with metadata before vectorization. This process ensures that the chatbot produces only accurate, traceable, and business-context-adapted responses.

Backend: Django REST Framework (API), Python

AI Orchestration: llama_index, LangChain, MistralAI

Vector Store: Qdrant (dense embeddings with MistralAIEmbedding)

Database: PostgreSQL

Task Queue & Cache: Huey, Redis

Frontend: Next.js / React with real-time response streaming

Objective: Explainable, secure AI anchored in trusted internal data

Drastic reduction in LLM hallucinations

Measurable improvement in response relevance (+90% on internal corpus)

Seamless integration into business tools, with supervision of sources used

This project illustrates my ability to design hybrid AI systems combining language models, vector search, and business logic to create reliable, contextualized, and auditable conversational assistants.

Development of a custom e-commerce site dedicated to product customization.

The application allows users to intuitively customize their items online, while respecting the customization restriction zones defined for each product.

A comprehensive administration interface allows site managers to:

Create and manage products via a dedicated dashboard.

Define customizable areas directly on the visuals.

Control the final rendering before it goes on sale.

This project highlights my skills in full-stack web development, dynamic media management, and user-oriented ergonomic interface design.