Llama Impact Hackathon

Start

Lapius, a new AI solution for medical coding autonomation

Lapius

JacopoC

Kushas Khadka

US hospitals and clinics are losing $125B every year due to delayed and inaccurate medical coding. While 8 out of 10 citizens receive medical bills with at least one error. Current solutions for both citizens and hospitals do not seem to be sufficient: even today, more than 50 percent of hospitals rely on the manual labor of medical coders. Although the large foundation models do not perform according to industry standards, we now propose a new framework, which combines multi-agent architecture with a RAG pipeline. At present there are two main functionalities: 

1. Extrapolate via Llama 3.2 the content of the codes of a medical receipt (e.g., an itemised bill)
2. Assign ICD-10 codes to a clinical note, explaining them and offering alternatives. 

The solution outperforms some of the solutions currently on the market. 

Lapius is a powerful Multi-agent application that empowers medical coders and patients to extract medical codes (ICD-10-CM) from medical records or explain the meaning of codes found in medical bills or EOB. 

"Lapius team on Llama Impact Hackathon Hackathon"

Team Idea

Submission