Boufia Oussama

Key Skills

Software

Appen

Mercor

OneForma

Telus

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Label Types

Bounding Box

Data Collection

Question Answering

Segmentation

Freelancer Overview

I am a detail-oriented AI Data Specialist and Linguistic Annotator with extensive experience contributing to AI training projects for platforms such as TELUS International, MagicData, DataPlus, and Ocean Speech. I specialize in data annotation, RLHF evaluation, audio transcription, and multilingual dataset optimization across Arabic (including North African dialects), French, and English. My work focuses on delivering high-accuracy labeled datasets for natural language processing (NLP) and computer vision models, consistently maintaining quality standards above 98% accuracy. With a strong background in linguistic quality assurance and SEO-driven content optimization, I excel at evaluating AI-generated responses for factual accuracy, cultural relevance, and instruction adherence. I am experienced in bounding box annotation, speech-to-text transcription with timestamping, and structured model evaluation workflows. My analytical mindset, attention to detail, and ability to interpret complex annotation guidelines allow me to contribute effectively to model fine-tuning and performance improvement projects.

Labeling Experience

Multilingual Named Entity Recognition (NER) & Text Classification Project

OneformaTextBounding Box

Annotated and classified multilingual text datasets in Arabic (MSA and North African dialects), French, and English for Named Entity Recognition (NER) model training. Tasks included identifying and tagging entities such as Person, Organization, Location, Date, and Numerical expressions according to strict annotation guidelines. Additionally performed text classification for intent detection, sentiment categorization, and topic labeling to enhance NLP model performance. Ensured high inter-annotator agreement (IAA) and maintained over 97% quality scores through rigorous self-review and adherence to project standards. Contributed to dataset validation and error analysis to improve model accuracy and reduce false positives in entity extraction tasks.

Annotated and classified multilingual text datasets in Arabic (MSA and North African dialects), French, and English for Named Entity Recognition (NER) model training. Tasks included identifying and tagging entities such as Person, Organization, Location, Date, and Numerical expressions according to strict annotation guidelines. Additionally performed text classification for intent detection, sentiment categorization, and topic labeling to enhance NLP model performance. Ensured high inter-annotator agreement (IAA) and maintained over 97% quality scores through rigorous self-review and adherence to project standards. Contributed to dataset validation and error analysis to improve model accuracy and reduce false positives in entity extraction tasks.

2024 - 2025

Arabic & French LLM Response Evaluation and RLHF Annotation

AppenTextSegmentation

Contributed to the training and optimization of large language models by evaluating and ranking AI-generated responses in Arabic (including North African dialects), French, and English. Tasks included assessing helpfulness, factual accuracy, cultural appropriateness, and linguistic fluency according to detailed annotation guidelines. Participated in Reinforcement Learning from Human Feedback (RLHF) workflows by comparing multiple model outputs and providing structured feedback to improve performance. Maintained a high agreement rate with quality reviewers and consistently met project KPIs for accuracy and turnaround time.

Contributed to the training and optimization of large language models by evaluating and ranking AI-generated responses in Arabic (including North African dialects), French, and English. Tasks included assessing helpfulness, factual accuracy, cultural appropriateness, and linguistic fluency according to detailed annotation guidelines. Participated in Reinforcement Learning from Human Feedback (RLHF) workflows by comparing multiple model outputs and providing structured feedback to improve performance. Maintained a high agreement rate with quality reviewers and consistently met project KPIs for accuracy and turnaround time.

2024 - 2025

AI Content Evaluation & Quality Assurance for LLM Optimization

AppenTextPrompt Response Writing SFT

Evaluated and rated AI-generated responses across Arabic, French, and English to improve large language model performance. Assessed outputs based on factual accuracy, coherence, instruction-following, harmful content detection, and cultural appropriateness. Conducted structured error analysis to identify hallucinations, bias, and logical inconsistencies in model responses. Performed safety-focused review (Red Teaming tasks) to test model robustness against unsafe prompts and edge cases. Maintained strict adherence to annotation guidelines and achieved consistently high quality review scores while handling high-volume datasets under tight deadlines.

Evaluated and rated AI-generated responses across Arabic, French, and English to improve large language model performance. Assessed outputs based on factual accuracy, coherence, instruction-following, harmful content detection, and cultural appropriateness. Conducted structured error analysis to identify hallucinations, bias, and logical inconsistencies in model responses. Performed safety-focused review (Red Teaming tasks) to test model robustness against unsafe prompts and edge cases. Maintained strict adherence to annotation guidelines and achieved consistently high quality review scores while handling high-volume datasets under tight deadlines.

2023 - 2024

Education

Work History