case
study

data lake engine

AI PROCESSING
OF MEDICAL DATA AT SCALE

HealthLab’s Client needed to implement AI to process semi-structured medical encounter notes, in sets of ranging from tens of thousands up to millions, to determine quality of medical practice.

At a Glance …

Key Metrics

2.5M Records

18K Image Files

87 Sub-measures AI Scored

(in progress)

CHALLENGES

Traditional, non-AI approaches (like RegEx and asking the clinicians to use macros) were not measuring quality accurately enough.

The Client needed human-level capability but at scale and without the expense.

Deep technical and medical knowledge were needed for LLM prompt design.

SOLUTIONS

The solution leveraged “big” data pipelines that could parallelize inputs at scale, call branching chains of logic with Large Language Models (LLM), and allow rapid and frequent iteration as exceptions and gaps in the LLM performance were identified.

LARGE LANGUAGE MODELS

“BIG” DATA PIPELINES

MEDICAL INFORMATICS EXPERTISE

BENEFITS


1


2


3

SCALABILITY
Petabyte scale processing capability with massive parallelization

CAPABILITY
Near human level capability at a fraction of the cost at pennies per encounter note

FLEXIBILITY
New measures can be deployed rapidly using new and emerging AI models to ensure best-in-class cost.