News for hackers. Stuff that matters.

We built an AI to analyze 100 biomarkers, then realized it has no intuition

We spent the last year building an AI to analyze 100+ biomarkers (metabolism, toxins, organ function) and perform predictive analysis on disease risks. The goal was to build a "Whole-Body Intelligence System". +1

The tech worked, but the solution was incomplete. We realized that while AI is excellent at functional grading across 10 health domains , it lacks "latent knowledge"—the intuition a doctor gains from 50 years of clinical practice that isn't in textbooks or datasets. +1

So, we pivoted from a pure-AI SaaS to a hybrid model: AI processes the data to find the signal, and a dedicated "genius doctor" interprets the "soul" of the data to create the protocol.

We are trying to solve the "last mile" problem in preventive health: turning raw data into reversed chronic conditions

http://nostaviahealth.com

( 1 comment | on HN )
Re:We built an AI to analyze 100 biomarkers, then realized i... by aiiizzz on Thursday December 25, 2025 @01:56PM (#46386318)
> the intuition a doctor gains from 50 years of clinical practice that isn't in textbooks or datasets.

You mean their possibly incorrect bias?