GODDESS: Generating Operational Datasets Designed with Exploitable Synthetic Scenarios

Artificial intelligence/machine learning (AI/ML) algorithm and model development have two fundamental problems: data scarcity and data wrangling. Despite this being an age in which data are ubiquitous and voluminous, data of a different type or composition for the algorithmic or modeling challenges faced are useless. Even where data type and composition are non-issues, the oft-cited work to transform the data for use remains roughly 80% of the overall development effort. What if data scarcity and data wrangling were no longer problems? What if you could generate all the data you needed in the required format on demand? What if data wrangling efforts could be decreased to 1/8th of the prior statistic (from ~80% to ~10%)? In this Phase I proposal, Aptima proposes Generating Operational Datasets Designed with Exploitable Synthetic Scenarios (GODDESS): a tool that will answer those questions with a resounding yes. GODDESS will ensure meaningful, helpful, and viable output using state-of-the-art single- and multimodal generative AI techniques with scenario-based guidance and a flexible, extensible, future-proofed system architecture.