AI that holds up in production.

Evaluating document-AI providers for an extraction pipeline — comparing a dedicated OCR service, a general LLM, and a document-native model at the line-item level, not just on headline accuracy.

Writing up the case studies behind these systems — the architecture, the trade-offs worth defending, and the parts that broke.

Open to senior applied-AI and solutions-architecture roles, and to selective consulting engagements — preferably where the model is one tool among several, not the whole product.