Leke Agbaje

Key Skills

Software

Labelbox

Scale AI

Top Subject Matter

Search Engine Evaluation

UK Digital Culture

Language Models

Top Data Types

Text

Image

Top Task Types

Classification

Bounding Box

Entity Ner Classification

RLHF

Freelancer Overview

Freelance Content Rater (Remote). Brings 1+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Labelbox and Scale AI. Education includes Master of Science, Sheffield Hallam University (2024) and Bachelor of Science, Nottingham Trent University (2021). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Classification.

Labeling Experience

Generative AI Quality Analyst

ImageClassification

My work involved Image Classification to assess the photorealism and structural integrity of AI-generated image edits. I reviewed AI generated images to ensure they stayed true to real world details. This involved a deep dive into visual cues, which included identifying specific application versions or checking that branding and logos were rendered accurately. My goal was to catch those tiny inconsistencies that make an image look fake, ensuring the final output was both high quality and factually correct.

My work involved Image Classification to assess the photorealism and structural integrity of AI-generated image edits. I reviewed AI generated images to ensure they stayed true to real world details. This involved a deep dive into visual cues, which included identifying specific application versions or checking that branding and logos were rendered accurately. My goal was to catch those tiny inconsistencies that make an image look fake, ensuring the final output was both high quality and factually correct.

2025 - Present

Visual Data Annotator

LabelboxImageBounding Box

I handled the high precision side of image tagging, where I focused on complex layouts like gaming screens and app interfaces. My main goal was making sure every 'bounding box' was a perfect tight fit. This was especially important for tiny interactive icons or grid elements where you really can't afford to catch any of the background in the box. It takes a lot of focus to manually separate individual buttons in a crowded menu, but i made sure every single component had its own clean box. This level of detail is what really helps a model understand exactly where objects are and how they are spaced out in a digital environment.

I handled the high precision side of image tagging, where I focused on complex layouts like gaming screens and app interfaces. My main goal was making sure every 'bounding box' was a perfect tight fit. This was especially important for tiny interactive icons or grid elements where you really can't afford to catch any of the background in the box. It takes a lot of focus to manually separate individual buttons in a crowded menu, but i made sure every single component had its own clean box. This level of detail is what really helps a model understand exactly where objects are and how they are spaced out in a digital environment.

2025 - Present

Fact-Checking & Data Integrity Specialist

TextEntity Ner Classification

I manually fact checked AI responses to ensure they weren't just making things up. This involved digging into the fine details like verifying a specific historical date or a person's name was actually correct and cross-referencing them against reliable sources. By catching these 'hallucinations' early, I made sure the final text wasn't just well written, but completely trustworthy and factually solid.

I manually fact checked AI responses to ensure they weren't just making things up. This involved digging into the fine details like verifying a specific historical date or a person's name was actually correct and cross-referencing them against reliable sources. By catching these 'hallucinations' early, I made sure the final text wasn't just well written, but completely trustworthy and factually solid.

2025 - 2026

Expert AI Training Specialist

Scale AITextRLHF

I was responsible for the 'human' side of AI training. I basically judge how well models actually followed instructions and whether they were being truthful. I spend my time digging into complex responses to catch those sneaky mistakes, like when a model gets its history mixed up or hallucinates a date that doesn't exist. It wasn't just about picking a favorite; I had to follow some intense, multi-page guidelines to make sure my feedback was completely fair and stayed focused on facts and data integrity. Every time I chose one response over another, I made sure to write out exactly why, so the developers had a clear path to making the model better.

I was responsible for the 'human' side of AI training. I basically judge how well models actually followed instructions and whether they were being truthful. I spend my time digging into complex responses to catch those sneaky mistakes, like when a model gets its history mixed up or hallucinates a date that doesn't exist. It wasn't just about picking a favorite; I had to follow some intense, multi-page guidelines to make sure my feedback was completely fair and stayed focused on facts and data integrity. Every time I chose one response over another, I made sure to write out exactly why, so the developers had a clear path to making the model better.

2025 - 2026

Education

Work History