N.O.A.H

Identity
History
Truth(Feedback)
Identity Chek-In
Identity Setup
Landing Page

Inspiration

I was frustrated by how most “AI” tools around mental health, productivity, or human performance only react after something breaks.
Burnout, dropout, collapse — they are treated as sudden events, when in reality they are almost always the result of a slow, invisible drift.

At the same time, I noticed a deeper problem with current AI approaches:
they rely heavily on black-box models that guess, hallucinate, or give advice without accountability.

Part of my inspiration also came from someone I secretly loved and admired deeply in my first university year.
She was extremely aligned with what she wanted and moved through life in a structured, systematic way.
That discipline and clarity left a strong impression on me, even though things didn’t work out between us.

Her name was one of the inspirations behind the name N.O.A.H.

N.O.A.H was born from a simple question:
What if we stopped trying to “understand” humans, and instead focused on observing trajectories?

What it does

N.O.A.H is a predictive observation engine that detects early signs of human drift using weak signals over time.

Users log simple signals (mental state, work, health, focus, social), scored on a small scale.
N.O.A.H aggregates these signals, detects patterns of repetition and decline, and classifies the user’s current state as:

Stable
Fragile
Drift
Critical

It then projects risk levels at 7, 14, and 30 days.

N.O.A.H does not diagnose, advise, or motivate.
It simply answers one question:

“If nothing changes, where is this trajectory heading?”

How I built it

I built N.O.A.H around a deterministic core engine, not a language model.

The system is composed of:

A signal collection layer
A temporal aggregation and drift detection engine
A state classifier
A trajectory projection module

All decisions are rule-based, auditable, and reproducible.

LLMs are intentionally excluded from the core logic.
They are reserved for a future interpretation layer, where they can explain outputs — never decide them.

The backend is modular, with N.O.A.H Core isolated as an independent engine that can be tested, versioned, and audited on its own.

Challenges I ran into

Resisting the temptation to use LLMs everywhere
It’s harder to design clear logic than to generate text.
Distinguishing noise from real drift
Humans are inconsistent by nature. I had to carefully design rules that detect persistence, not momentary fluctuation.
Ethical boundaries
I had to be extremely precise about what N.O.A.H should not do: no diagnosis, no advice, no authority over the user.
Designing “cold” UX
Most products try to comfort or motivate. I deliberately chose restraint.

Accomplishments I’m proud of

Designing a predictive system without machine learning
Building a core engine that is:
- deterministic
- explainable
- falsifiable
Creating a clear separation between decision logic and AI interpretation
Producing a product that could realistically operate in sensitive domains
Avoiding hype while still delivering real insight