The primary objective was to curate a diverse dataset consisting of various user interactions with software tools.
The dataset is pivotal for enhancing AI's understanding of human-computer interactions, involving planning and execution to capture diverse user actions across multiple software applications.
We assembled a skilled and diverse workforce, including professionals from various domains such as research, technology, finance, engineering, and creative design, which ensured coverage of interactions across different tools and scenarios.
We created an in-house tool to capture actions and designed a process for data creation and audit at scale.
We successfully delivered a diverse dataset for Adept's action transformer model.