The problem
Medical records arrive as a mix of scans, lab tables, handwritten notes and PDFs, written in language most patients cannot follow. Clinicians lose time interpreting reports from outside their own specialty, and caregivers and researchers hit the same wall when working through large volumes of records.
What we built
A conversational AI assistant that reads uploaded medical reports and answers questions about them in plain language, producing a structured summary covering patient information, diagnostic findings, risk assessment and recommended next steps. It ingests CSV, PDF and image formats, supports English and Portuguese, adapts its tone to the reader, and is explicitly prevented from recommending specific medications.
The system runs on Google Cloud with dedicated processing so that no data leaves the provider's controlled environment. It integrates with EPIC as the first EHR target and supports FHIR, HL7 and ICD classification through an OAuth 2.0 API, with the design built to HIPAA, GDPR, LGPD and ANVISA requirements.
Outcome
- Live in 2026 with the provider
- Document ingestion, structured summarisation and multilingual question-answering in production
- Compliance-ready by design across four regulatory regimes, with clinical safeguards built into the assistant's behaviour
Customer details available under NDA. Talk to us about document and conversational AI in regulated healthcare.