Marysia Winkels

Driving data quality and diversity initiatives for post-training datasets of Cohere's flagship models, Command A and Command R/R+.

Research project on representation engineering in collaboration with Center for AI Safety.

PyData is a community of data scientists and developers that share and learn about developments of data analytics tools and methodologies, mostly in Python. The committee organizes monthly meetups and a yearly conference hosting 300+ people.

Designed various data science curricula and subsequently delivered over 1,000+ hours of training to data professionals for Xebia Academy. Topics covered included analytics translation, data analysis, (advanced) machine learning, MLOps, coding best practices, data visualization and storytelling, and deep learning fundamentals and practical applications.

Leading cross-functional data science initiatives spanning multiple product areas across Vattenfall B2C, driving both technical direction and strategic decision-making.

Developed models for the detection and quantification of lung nodules in 3D CT scans.

Thesis on geometric deep learning applied to the medical domain, specifically early lung cancer detection. This work resulted in a direct 10x increase in data efficiency.

Exploratory internship on meniscal tear detection in MRI scans and the automatic parsing of medical reports in Dutch.

Teaching Assistant for various second year BSc Artificial Intelligence courses, such as Calculus & Statistics, Computer Vision, Natural Language Models & Interfaces, Knowledge Systems, Cognitive Psychology, Webprogramming & Databases.

Thesis subject: data efficient deep learning on biomedical image data with group-equivariant neural networks Internships at Microsoft and Aidence (now DeepHealth).

Thesis subject: fair division and resource allocation. Honours programme, minor in Language Philosophy