“The best way to send knowledge from one place to another is to wrap it up in a person”. This famous quote by Robert Oppenheimer depicts the importance of individuals in acquiring and transferring knowledge. Among individuals acting as vectors for knowledge transfer, PhD graduates, a highly educated labor force sitting at the knowledge frontier, are in a privileged position to spread knowledge related to novel technologies. One such technology is today Artificial Intelligence (AI), which governments in all advanced countries support through large investments in universities and public research organizations. The AI-Edu project focuses on France and seeks answers to three research questions: (i) Where are French PhD students trained in AI? (ii) What are the determinants of students’ AI training during the PhD period? (iii) How does students’ training in AI affect their careers, productivity, and professional networks? AI-Edu uses microdata at the individual level, including the whole population of French PhD graduates in STEM from 2000 to 2022. For each PhD graduate, we collect thesis, publication, and patent information. We assess the AI content of each thesis by analyzing the text of its title and abstract using neural network algorithms for text analysis. We also interview PhD students and supervisors to shed light on the micro-mechanisms leading to AI training during the PhD and driving AI training impact. Policymakers and the French government benefit from AI-Edu results by having a complete assessment of drivers and the impact of AI training for the entire population of PhD graduates over time and across scientific and technological fields. AI-Edu also contributes to the labor market literature with an unprecedented analysis of the supply of AI-trained individuals and to the education literature studying PhD students’ training outcomes.