AI + Audit · Product Innovation
National product-innovation challenge by Deloitte China. I advanced from the AI+Audit preliminary track (Top 5 individual) to the national finals and led the team on the technical side. We designed and shipped a working prototype for an AI-enabled audit workflow: sampling → document understanding → checks → workpapers, focusing on reliability and explainability.
Cross-functional contests reward clarity more than flash. Owning the full stack taught me to translate audit requirements into testable product behaviors: objectives → risks → controls → evidence. Two choices paid off: (1) a modular design that tolerates imperfect inputs with graceful fallbacks; (2) surface-level guardrails and reviewer overrides so auditors stay in control. Being the only technologist also reinforced a habit—show the smallest valuable loop first (upload → check → workpaper), then iterate on accuracy and speed.