
The client had a 100k+ archive of informed consent records linked to high-value clinical biosamples, but reuse permissions were buried in complex legal and clinical language.
Hybrid Mind helped turn that static archive into structured, Al-searchable consent data - giving scientists and governance teams a faster, more defensible way to identify which samples could be reused for future research.
Results: Hybrid Mind turned a 100,000+ record consent archive into a searchable, governedworkflow for research-ready biosample identification. Project reporting showed review effort moving from hours per document to minutes, while model accuracy improved from a 60-70% POC baseline towards 90%+ performance on key outputs.

Claude LLM via AWS API | Python | Domino | Few-shot prompting | Prompt chaining | Rule-based post-processing | Ground-truth validation | Confidence scoring | Business-rule codification

