Case study
AI/LLM to Extract Information from Medical Records
The client is an Israeli entrepreneur in the medical field who wanted to ensure that the AI solution he needed could be developed. He had tight deadlines and additional functional requirements.
About the client
The client is an Israeli entrepreneur in the medical field who wanted to ensure that the AI solution he needed could be developed. He had tight deadlines and additional functional requirements.
Background and problem
The client needed a way to extract important information from medical records. They required an AI solution that could work on-site without needing an internet connection. Additionally, they needed a tool that could convert text descriptions into ICD-9 codes.
Solution
Our team designed and completed a project to extract critical information from medical records, focusing on Biomarkers, Stage, Grade, and Histologic diagnosis. The application operates on-premises, without internet connectivity, and with minimal hardware requirements. The tool also allows users to obtain a code from the ICD-9 classifier based on a textual description.
Our team successfully designed and completed this project. The implementation involved a two-step process: initially leveraging a large language model (LLM), followed by fine-tuned, trained models for enhanced precision.
More cases
See all cases-
AI Platform for Anomaly Pre-Screening and Telemedicine Monitoring
Our cient is the World leader in FDA-cleared portable ultrasound for home use. We built an Azure cloud platform with AI-powered anomaly detection, real-time video guidance, and remote device control reducing in-clinic visits for pregnancy monitoring.
-
Automated Insurance Verification for a Stealth HealthTech Startup
HealthTech startup automating insurance verification for medical practices. Our AI platform runs 10,000+ benefit checks a month, cutting front-desk time from 8–11 hrs/week to 1–2. When a name is misspelled or a card expires, the system texts the patient for a photo of their insurance card, corrects the information, and re-runs the check. Staff step in only for cases the platform can't resolve on its own.
-
Health platform for IRIDIUM
US-based continuing medical education company. We built AI patient avatars simulating real doctor visits, with LLM-powered dialogue analysis integrated into LearnWorlds and FreeCME. When a doctor orders labs or uncovers a symptom, the avatar updates its own medical chart in real time. No manual input, no static scripts.
Let’s discuss your project
Book a free consultation with our healthcare AI team.