Consultant.e.s - Stage de fin d'études - Nantes - CIO - IA responsable : quels KPI et outils pour suivre et implémenter explicabilité, biais, censure ?

Define types of conversational biases in generative AI chatbots and benchmark detection tools like IBM AI Fairness 360. Map techniques for bias reduction across the chatbot lifecycle, including fine-tuning, prompt filtering, and RLHF. Establish KPIs for response diversity, equity, and stereotyping to guide decision-makers. Draft a best-practices guide and decision-support tool for project teams and stakeholders. This internship supports Wavestone's emerging expertise in responsible AI governance, specifically for clients in regulated sectors like finance and public administration. The role contributes to building a new service offering focused on ethical chatbot design.