HiWi: Medical Guideline Collection and Pre- processing for AI-Powered Decision Support

Stelle ist zu besetzen ab: 01.09.2025

Beschreibung

Introduction:
Clinical guidelines are fundamental for evidence-based medicine but are often long and complex
for clinicians to apply efficiently. Recent AI advancements, particularly with Large Language
Models (LLMs), offer a powerful solution. Previous researches have shown that visual
guidelines, like flowcharts, can be converted into machine-readable format for LLMs. The goal of
this HiWi position is to systematically collect and process a diverse set of medical guidelines.
This prepared dataset will enable future research into AI systems that can help clinicians make
more accurate and more personalized treatment decisions.

What You'll Be Doing (and Learning!):
• Research & Collect: Identify and systematically acquire publicly available medical guidelines
from various specialties and reputable sources.
• Extract & Convert: Utilize and evaluate state-of-the-art models to extract text, tables, and
flowcharts from PDF documents. A key task is converting this extracted information into
structured, machine-readable formats like JSON.
• Structure & Clean: Process the raw extracted data to create a clean, organized, and analysis-
ready dataset. This dataset must be specifically formatted for effective training of Retrieval-
Augmented Generation (RAG) models

Who We're Looking For:
• Interest in the topic of computer vision and healthcare
• Python programming skills and knowledge of PyTorch are desirable

Further Information:
• Start date: 01.09.2025
• Hours: 30-40 hours/month (initially for 3 months, extension is possible and desired)
• Contact: M.Sc. Jiale Wei

Please feel free to contact me(jiale.wei@kit.edu) with your CV and Transcript of Records.