Arnold Obonyo

Key Skills

Software

Telus

Other

Top Subject Matter

Text Annotation and Content review

Prompt Response Ranking

Data Categorization and Cleanup

Top Data Types

Text

Image

Document

Top Task Types

Classification

Bounding Box

Polygon

Point Key Point

Freelancer Overview

AI Data Annotator. Core strengths include Telus and Other. Education includes Bachelor of Computer Science, Multimedia University (2022). AI-training focus includes data types such as Text and Image and labeling workflows including Classification and Bounding Box.

Labeling Experience

AI Data Annotator

TelusTextClassification

As an AI Data Annotator at Telus Digital, I handled high-volume annotation tasks, including text classification, response ranking, and reasoning chain annotation. I consistently implemented complex annotation guidelines on factuality, tone, safety, and user intent. My role included quality assurance, frequent calibration meetings, and examining model reasoning chains for errors. • Processed over 1,500 annotation tasks and reviewed 500+ peer annotations weekly • Maintained a 98% client QA acceptance rate and 95% first-time accuracy on gold set searches • Identified hallucinations, logical gaps, and error rates in model outputs, reducing model errors by 15% • Cleaned and formatted raw model results for SFT and RLHF pipelines

As an AI Data Annotator at Telus Digital, I handled high-volume annotation tasks, including text classification, response ranking, and reasoning chain annotation. I consistently implemented complex annotation guidelines on factuality, tone, safety, and user intent. My role included quality assurance, frequent calibration meetings, and examining model reasoning chains for errors. • Processed over 1,500 annotation tasks and reviewed 500+ peer annotations weekly • Maintained a 98% client QA acceptance rate and 95% first-time accuracy on gold set searches • Identified hallucinations, logical gaps, and error rates in model outputs, reducing model errors by 15% • Cleaned and formatted raw model results for SFT and RLHF pipelines

2024 - 2025

LLM Safety & Harmful Content Labeling Pipeline

OtherTextClassification

I participated in a language model safety project, annotating AI-generated responses into 12 security categories for model fine-tuning. My work followed a 45-page guideline and involved classifying content as Safe, Unsafe, or Edge case with detailed breach type labels. I ensured high inter-annotator agreement while carefully logging challenging cases for calibration. • Categorized responses addressing hate speech, self-harm, explicit content, and unlawful activities • Applied complex guidelines across thousands of AI outputs • Achieved 99% inter-annotator agreement through weekly sessions • Contributed to a dataset reducing unsafe outputs by 22% in production safety classifiers

I participated in a language model safety project, annotating AI-generated responses into 12 security categories for model fine-tuning. My work followed a 45-page guideline and involved classifying content as Safe, Unsafe, or Edge case with detailed breach type labels. I ensured high inter-annotator agreement while carefully logging challenging cases for calibration. • Categorized responses addressing hate speech, self-harm, explicit content, and unlawful activities • Applied complex guidelines across thousands of AI outputs • Achieved 99% inter-annotator agreement through weekly sessions • Contributed to a dataset reducing unsafe outputs by 22% in production safety classifiers

Not specified

Image Bounding Box & Polygon Annotation Project

OtherImageBounding Box

In an image annotation project, I drew bounding boxes and used polygon segmentation for vehicles, pedestrians, and traffic signs in street view images. My tasks included creating attribute tags such as color, direction, and obstruction estimates. I maintained high annotation quality metrics and documented difficult edge cases for the team. • Annotated more than 1,200 bounding boxes and polygons for various objects • Focused on difficult cases such as partially hidden items and shadows • Consistently achieved an IoU score above 0.85 on validation image sets • Built a portfolio-worthy dataset for use in multimodal AI projects

In an image annotation project, I drew bounding boxes and used polygon segmentation for vehicles, pedestrians, and traffic signs in street view images. My tasks included creating attribute tags such as color, direction, and obstruction estimates. I maintained high annotation quality metrics and documented difficult edge cases for the team. • Annotated more than 1,200 bounding boxes and polygons for various objects • Focused on difficult cases such as partially hidden items and shadows • Consistently achieved an IoU score above 0.85 on validation image sets • Built a portfolio-worthy dataset for use in multimodal AI projects

Not specified

Education

Work History