Staff Research Scientist at Sakana AI’s RSI Lab, studying AI systems and how people learn with them. Her earlier work includes Google DeepMind, Cognimates and family-centered AI-literacy research.[1][5][3]
Start with Kids teach AI a little humanity with Cognimates →
Education
Master’s, Media Engineering for Education (EUROMIME)
Sep 2008 – Sep 2010
[8][9]
Bachelor of Communication, Public Relations and Communication
Completed 2005
[10]Links
Website University research history Résumé
X · Stefania_druga Staff Research Scientist, RSI Lab · Sakana AI
Career
Druga founded Hackidemia in 2012 after earlier work in Google’s Search Quality team. At MIT she developed Cognimates, connecting programming with children’s exploration of AI. She later completed a doctorate at the University of Washington and worked as a research scientist at Google DeepMind before joining Sakana AI. Her current research examines self-improving systems, memory and collaboration.[2][10]
Contributions
Let children learn by training and questioning a system
Cognimates extends Scratch-based creative programming with AI-learning activities for children. Participants can build projects, train models and interact with embodied agents. Druga’s account connects the platform with learning by teaching: making a system respond requires children to consider what examples and instructions it has received.[2]
Study the roles parents play in learning about AI
Her coauthored family-literacy study designed eleven activities covering image classification, object recognition, voice assistants and unplugged co-design. Eighteen children and sixteen parents participated in a five-week in-home study. The analysis examines how parents support and collaborate with children, then offers design recommendations for family-centered learning resources. The activities make AI concepts discussable through shared experiments.[3]
Tools, writing & talks
Writing, research & talks
Start here
Kids teach AI a little humanity with Cognimateswww-prod.media.mit.eduDruga’s account of building an open-source AI-literacy platform and the educational experiences behind it.
Start with the Cognimates article for concrete learning activities and the reason for extending creative programming into AI. It requires no research background; the family study supplies the later empirical methods and recommendations.
Author and Cognimates creator[2]On a literature review where every paper already fitted in the context window, adding a memory harness changed nothing but cost. That negative result marks the boundary, before the harness becomes decisive on long-horizon tasks.
Catalog workOriginal source- Family as a Third Space for AI Literaciesstefania11.github.io
A study of parent-child learning activities and parents’ roles in developing AI literacy.
Coauthor[3]