Laafou Hamza

Key Skills

Software

Labelbox

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Text

Top Label Types

Prompt Response Writing SFT

RLHF

Freelancer Overview

With a strong background in graphic and UI design, I bring a keen eye for detail and accuracy—qualities essential for high-quality data labeling and AI training data roles. My experience working with industry-standard tools like Photoshop, Illustrator, Figma, and Adobe XD has honed my ability to interpret visual data, ensuring precise annotation and categorization. Having worked on branding, editorial, and advertising projects, I am adept at understanding complex visual information and maintaining consistency across datasets. I am passionate about supporting AI development by delivering reliable, well-structured training data for computer vision and user interface applications.

Labeling Experience

LLM Alignment Specialist - Preference Ranking & Safety

Scale AITextRLHF

Performed high-complexity data annotation and quality assurance to improve the reasoning and safety capabilities of a Tier-1 Large Language Model. Scope: Focused on multi-turn dialogue evaluation and Supervised Fine-Tuning (SFT) for coding and creative writing prompts. Tasks: Ranked model-generated responses based on strict dimensions: Truthfulness, Helpfulness, and Harmlessness. Wrote "Golden Responses" to serve as ground-truth data for model training. Project Size: Processed and audited over 2,500+ complex prompt-response pairs. Quality Measures: Maintained a consistent Quality Audit (QA) score of 97% or higher. Adhered to complex, evolving style guides and utilized "Chain of Thought" reasoning to justify ranking decisions.

Performed high-complexity data annotation and quality assurance to improve the reasoning and safety capabilities of a Tier-1 Large Language Model. Scope: Focused on multi-turn dialogue evaluation and Supervised Fine-Tuning (SFT) for coding and creative writing prompts. Tasks: Ranked model-generated responses based on strict dimensions: Truthfulness, Helpfulness, and Harmlessness. Wrote "Golden Responses" to serve as ground-truth data for model training. Project Size: Processed and audited over 2,500+ complex prompt-response pairs. Quality Measures: Maintained a consistent Quality Audit (QA) score of 97% or higher. Adhered to complex, evolving style guides and utilized "Chain of Thought" reasoning to justify ranking decisions.

2024 - 2025

Education

Work History