Antoine de Reviers

Building LLM-powered GenAI solutions. Leveraging advanced techniques including: - Structured data extraction - Classification - Enhanced search - Retrieval-Augmented Generation (RAG)

At Scenario, a startup developing generative AI for the gaming industry, I contributed to key projects including: - Developing the inference backend: industrialization of open-source Diffusers pipelines with continuous integration of the latest features around Stable Diffusion. - Developing the training backend: industrialization of diffusion models fine-tuning algorithms (DreamBooth, LoRA). - Building various stand-alone features with FastAPI and SageMaker endpoints. - Creating Gradio demos with deployment on HuggingFace spaces. Env: AWS, Python (PyTorch, Diffusers, FastAPI, Pydantic), Docker, CI/CD

Focused on advancing SAR image recognition under constraints: - Conducted research and state-of-the-art analysis on model development. - Built comprehensive radar databases for enhanced model training. - Developed robust pipelines for radar and image processing integration using Matlab and Python. - Implemented advanced recognition models capable of handling open-world classification and out-of-distribution detection.

Engineered machine learning solutions for automating part sorting in production: - Developed and deployed computer vision models integrating simulated data with real images. - Built a custom image annotation tool and established a photo setup integrated within the production line. - Developed a comprehensive web app for remote control and data management, enhancing operational efficiency.