yu s

Développement d’une plateforme IA on-premise - Backend : FastAPI, Docker, Kubernetes, GitLab CI/CD - Implémentation : Authentification, monitoring, tests unitaires - IA : Optimisation d’un pipeline RAG (Hugging Face embeddings & reranker) pour un gain mesurable en qualité de réponse NLP

Designed and implemented an NLP system integrating Whisper AI, GPT-4, and LLaMA 3 in Python. ◦ Management of startup competitions (3 prizes won) ◦ collaboration with external partners for the development of AI conversational assistants.

Designed and implemented a Retrieval-Augmented Generation (RAG) architecture to process and reason over dynamic, real-time sensor data for intelligent monitoring and contextual decision-making. Developed a hybrid system integrating embedded LLMs (LLaMA 3) with a custom vector store powered by FAISS, enabling low-latency retrieval of time-sensitive data snapshots. Implemented prompt engineering and chaining strategies using LangChain, allowing the agent to adapt queries based on evolving environmental signals and historical memory.

Bioimage Processing and Analysis, Machine learning, Computer vision, Classification algorithms, Interdisciplinary collaboration, Technical documentation and presentation