# I Built an AI Judge That Doesn't Decide Who Wins.

> Source: <https://dev.to/dikshit_garg/i-built-an-ai-judge-that-doesnt-decide-who-wins-58h9>
> Published: 2026-10-04 18:34:39+00:00

*This is a submission for the [Hacktoberfest Weekend Challenge: Build for a Friend](https://dev.to/challenges/hacktoberfest-weekend-2026-10-01)*

My friend participates in hackathons frequently and builds strong projects, but she had an important concern:

**How can you know that judges are applying the same rubric consistently?**

A project can receive different scores from different judges even when everyone is evaluating the same criteria.

So I built **FairJudge AI**.

FairJudge is an AI-assisted hackathon evaluation system designed to make judging more **transparent, evidence-based, and consistent**.

It can:

But there is one important thing FairJudge **doesn't** do:

**It does not decide who the objectively correct winner is.**

The final decision always remains with human judges.

Hackathon judging involves multiple judges evaluating projects across several criteria.

Disagreements can happen because:

Instead of building another AI that simply says:

"Project X should win."

I wanted to build something more transparent:

**Why was this score given? What evidence supports it? What evidence is missing? Where do judges disagree?**

That became **FairJudge AI**.

🎥 **Demo Video:**

[https://claude.ai/artifact/9JLZQKa7C33ADuJg2Qk5wC](https://claude.ai/artifact/9JLZQKa7C33ADuJg2Qk5wC)

The demo shows:

The core workflow is:

```
text
Project + Rubric
       ↓
Evidence-based AI Evaluation
       ↓
Evidence Validation
       ↓
Deterministic Python Scoring
       ↓
Multiple Judge Comparison
       ↓
Disagreement Detection
       ↓
Human Review
```