Large Language Models (LLMs)

More information to come!

Overview

Unstructured clinical text accounts for over 80% of health data. Traditional natural language processing (NLP) methods lack contextual flexibility and domain generalization. Large Language Models (LLMs) enable advanced capabilities: summarization, classification, translation, reasoning.

Ollama

Ollama is an open-source framework for managing, running, and serving large language models locally on host hardware.

Prompt Engineering at Scale

• Standardizes prompt templates and variables

• Runs multiple prompts iteratively for evaluation or comparison

• Logs inputs/outputs for reproducibility and auditability

Input

• Imports both structured data (like spreadsheets/CSVs) and standardized healthcare files (like FHIR patient records)

• Cleans and extracts raw narrative text from unstructured sources, such as messy clinical notes

• Clinical tables and free text can be combined in prompts.

Output

• LLM responses parsed and stored programmatically.

• Results may be written to data frames or databases for downstream use.