← All case studies
healthcareDocument AI
600-bed hospital network
AI pipeline ingests handwritten patient forms, extracts fields, writes to Epic EHR. Intake dropped from 40 min to 8 min per patient.
80%
Faster intake
14M+
Docs processed
99.4%
Accuracy
12 wks
To production
The challenge
40 minutes per patient, just for paperwork.
A 600-bed hospital network across 3 locations was drowning in paper. Every new patient meant handwritten forms manually keyed into Epic EHR. Average intake: 40 minutes.
What we built
OCR + LLM pipeline that handles messy forms.
Three-stage pipeline: scan, extract, validate, write. High-res scanning, OCR with post-processing, LLM interpretation, validated writes to Epic via HL7 FHIR API.
The results
8 minutes. Not 40.
Patient intake dropped from 40 to 8 minutes. 14M+ documents per year with 99.4% accuracy. Paid for itself in 4 months.
TECH
Python, Tesseract, Claude API
INFRA
AWS, Docker, Redis
INTEGRATION
Epic EHR, HL7 FHIR
“We went from 40 minutes of data entry per patient to 8.”
VP of Clinical Operations · 600-bed hospital network
Got a similar challenge?
We'd love to hear about it. First call is free.