Tuterangi Cordette

Design and development of a RAG tool dedicated to automatic report writing. Study of retrieval improvement methods to optimize RAG performance. Study of MCP methods applied to AI agents. Study and implementation of RAG response evaluation systems (Deepeval, HuggingFace Evaluate). Configuration of an Oracle RED2 server.

Research work on AI-based skin lesion detection methods in a project aimed at preventing skin cancer. Design and training of a U-Net neural network for lesion segmentation on medical images. Implementation of a detection method using a Sobel filter to extract and analyze lesion contours. Development of a statistical tool applied to images to quantify and evaluate the characteristics of detected lesions.

Development of the backend for an internal chatbot for IMSA during the La Mêlée hackathon. Implementation of retrieval-augmented generation (RAG) pipelines to improve knowledge access and response accuracy. Integration of LangChain with the OpenAI API to design modular and scalable conversational flows. Experimentation with routing techniques and agent-based architectures to handle complex user requests. Collaboration within a multidisciplinary team to deliver a functional prototype under tight deadlines.