Guardoc reports 46% fewer documentation errors running clinical PDFs through a two-tier Amazon Nova pipeline
AWS Machine Learning Blogmedium signalA July 27 AWS post walks through Guardoc Health's seven-stage long-term-care documentation pipeline: Textract extraction, chunking at clinical boundaries, Titan Text Embeddings V2, patient-partitioned DynamoDB storage, type/recency pre-filtering, k-NN retrieval, then Nova for final classification with source traceability. The cost tiering is the interesting design choice — Nova 2 Lite handles text review and filtering, and only the final multimodal reasoning over PDF layout, handwriting and visual context escalates to Nova Pro. Guardoc reports 46% fewer documentation errors, 70% fewer audit fines, 65% lower litigation exposure, $400K+ annual ROI per facility, and a 74% reduction in hospital transfers per 100 admissions in one deployment; no latency or token-cost figures are published.
↳ Follow the thread