How proudP Trains Its AI Model for Urine Flow Analysis
- Andy Jung, PhD
- Jan 23
- 2 min read
Updated: Feb 23
One of the most common questions I receive is,
“If proudP analyzes the sound of urination… how did you even train the first AI model?”
The short answer is simple, but not easy:
We had to create a dataset that didn’t exist anywhere in the world.
The Core Challenge: No Simulator, No Synthetic Shortcut
In many AI domains, early models can rely on:
Simulators
Synthetic data
Controlled, repeatable inputs
Urine flow doesn’t work that way.
It’s a highly non-linear, turbulent physical process, and synthetic data simply couldn’t capture this complexity.
The Data We Needed to Build Our At-Home Urine Flow Test App
To build a predictive model, we needed paired data captured simultaneously:
The sound of real urination in a normal bathroom
Clinical-grade urine flow-rate data, the same type used in hospitals
The problem is that these two things normally happen in completely different environments.
We couldn't match the data by having someone urinate twice since bladders aren’t machines.
Our Solution: A Very Un-Digital MVP
Our first MVP didn’t look like digital health at all.
It was:
A ceramic toilet
Mounted on a specially designed precision scale
Engineered to capture sound and true flow-rate simultaneously
This setup allowed us to record:
Moment-by-moment acoustic data
Matched exactly to real flow-rate measurements
This gave us the paired, moment-by-moment data needed to train our earliest AI model.
Testing Across Real-World Variables
Once the core setup worked, we expanded aggressively.
We tested across real-world variables:
Bathroom echo
Background noise
Water level differences
Smartphone types and microphones
The dataset started in-house (yes, the founding team’s data is in there 🚽), then expanded into clinical studies.
Only after laying this groundwork could we build proudP’s AI - the one that is now FDA-listed and clinically validated.
Understanding the Importance of Accurate Data
Accurate data is crucial for any AI model. It ensures that the predictions made are reliable. In our case, the sound of urination needed to be captured in various settings. This way, we could account for different factors that might affect the results.
Why Sound Matters
The sound of urination is not just noise; it carries vital information. By analyzing these sounds, our AI can estimate urine flow rates effectively. This is a game-changer for men looking to monitor their urinary health from the comfort of home.
ProudP: Your At-Home Urine Flow Test Solution
proudP is an at-home urine flow test app that uses acoustic AI to measure urine flow with your smartphone—no clinic visit required.
Learn more about how to use proudP in the post below.
