Scott W.

Key Skills

Software

Other

Mercor

Top Subject Matter

Math Domain Expertise

Physics Domain Expertise

Computer Programming

Top Data Types

Text

Computer Code Programming

Top Task Types

Text Generation

Question Answering

Fine-tuning

Evaluation/Rating

Computer Programming/Coding

Function Calling

Freelancer Overview

I've been actively working as a freelance AI trainer and response evaluator on Outlier focused primarily on STEM-heavy domains: mathematics, physics, programming in C++, and formal logic and puzzle problems. My standard workflow is to independently solve each problem to establish ground truth — often writing Python scripts for computational verification — before scoring model responses against a structured rubric covering instruction following, accuracy, verbosity, format, and writing style, followed by a winner verdict and itemized error flags. This ground-truth-first approach reliably catches subtle reasoning failures, fabricated citations, and silently incorrect derivations that surface-level review misses.

Labeling Experience

Outlier

TextEvaluation/Rating

I first determine if a STEM-related prompt can be evaluated. Once verified, I am presented two responses from different LLMs and I choose which response is better. I provide a detailed justification for my decision in each case.

2026 - Present

Education

Work History