Figma

This post was last edited by an anonymous forum user on 2025-09-27 20:06. Five rounds: one coding round, one ML system design round, one ML model design round, and one behavioral questions (BQ) round. The following content requires a score higher than 188. You can already view it. Coding: Completely forgettable. ML system design: Recommendations for Figma assets/templates. Similar to Facebook's newsfeed resys, this is sufficient. ML model design: Suppose Figma needs to create a prompt-to-design VLM. To simplify the question, assume the design only requires a conceptual diagram. It covers all aspects of model end-to-end, especially how to crawl and label data, and how to perform data augmentation. BQ: Asked many questions about cross-functional collaboration. Project deepdive: Just a regular project deepdive. It's best to talk about something related to computer vision (CV), otherwise the interviewer might not understand the concepts and could misunderstand the scope. Finally, I was challenged due to insufficient years of experience, possibly at the org level 0-1, indicating insufficient experience.