All dev work

Outlabs

Diagnostic algorithms

nmillrr

Role:

Developer

Date:

Nov 2025

Preview

Github

Building an open-source library of biomarker models for clinical diagnostics education

As a pre-med student, I worked in a few different offices and clinics that used lab tests. When I started looking into how much these test cost for patients, I was surprised to learn how much money could be saved if there was a different business model available. Patients without insurance might pay hundreds of dollars for basic panels, despite the clinical cost often only being a few dollars.

In the world of digital products and services, most users skip over the lengthy terms and conditions agreements (alternatively called 'EULA's, or end-user license agreements). These agreements may contain complex legal jargon, thousands of words, and other obstacles that make a thorough read-through unrealistic for the average user.

Termes delivers an accessible solution to this problem: a browser extension that helps users better understand what they are agreeing to. There are three major parts to Termes: a grading scale (inspired by ToS;DR), a highlight of any violation of rights, and a high-level summary of the terms.

Since the unit cost of these tests are so low, patients are really paying for the labor and overhead of the diagnostic labs. However, from working in low-resource clinics, I learned that expensive lab equipment can often be sourced secondhand through corporate or small business sell-offs online. But to truly democratize diagnostics, clinics also need biomarker models, which can often be held as proprietary algorithms to specific labs.

Outlabs is a data science project I'm building to provide clinically-validated Python models that estimate biomarkers from blood tests. By helping make diagnostics more accessible, I'm hoping not only to save patients money, but also educate rural clinics, low-resource hospitals, and global health organizations worldwide.

This project is in progress, but you can test the models and check out the Github repo here. Currently working on adding more models to the list and more rigorous clinical validation.