Industrial project built in partnership between ESPRIT and Vital Laboratoire. Developed the full data infrastructure powering an AI conversational avatar for pharmaceutical product support. Built end-to-end web scraping workflows to collect product data from SPMC databases, Wikipedia, and institutional sources using BeautifulSoup, Requests, and Scrapling - handling both modern and legacy websites with dynamic JS content. Structured and normalized all extracted data into clean knowledge bases feeding the AI avatar's conversational engine.