The impact of data errors on the output of AI models is difficult to predict and quantify, even though such errors pose real risks for regulatory compliance. This project therefore focuses on supporting non-technical users in validating and improving the quality of their data. Users should be able to formulate data quality rules in natural language. We develop a data-driven approach that leverages these rules to help domain experts refine data quality constraints and perform “stress tests” on downstream AI models. This project is particularly well suited for someone with a strong background in data engineering and an interest in collaborating with domain experts.
