abstract
- Investigates systematic ways to debug and validate scientific workflows. A researcher survey characterizes existing debugging practices; an input-reduction framework helps isolate data responsible for unexpected outcomes; and contract-driven design formalizes requirements and guarantees for workflow tasks. The thesis also explores automated contract generation with the SANA pipeline, aiming to make workflow diagnosis and validation less dependent on ad hoc manual work.