Muhammad A.

Key Skills

Freelancer Overview

As a backend engineer and statistician, my experience with AI training data focuses on high-complexity data curation, algorithmic verification, and evaluation of advanced reasoning chains. I bring a deep, practical understanding of model mechanics from building a custom GPT transformer from scratch using PyTorch, which required engineering tokenization pipelines, handling large-scale unstructured training datasets, and writing custom validation loops to assess cross-entropy loss. My background in 300-level statistics provides me with specialized expertise in data modeling, distribution theory, and numerical analysis, enabling me to analyze, label, and optimize complex datasets with high precision. What sets me apart is my ability to rigorously evaluate and benchmark AI outputs for technical accuracy. Beyond engineering distributed systems, I have over a year of experience teaching and mentoring junior statistics students, where I regularly design complex problem sets and code labs to translate mathematical theory into validated Python and R scripts. This combination of educational leadership, statistical expertise, and strong programming skills allows me to effectively audit multi-turn logic, identify subtle code anomalies, and generate high-quality adversarial training data that directly improves LLM performance.

Education

Work History