Mustapha Alude

Key Skills

Software

Clickworker

Labelbox

Label Studio

Mercor

Mindrift

Toloka

Telus

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Action Recognition

Bounding Box

Classification

Emotion Recognition

RLHF

Freelancer Overview

Analytical and detail-oriented AI professional with over two years of experience evaluating and refining generative AI systems. Skilled in analyzing model outputs for logical reasoning, factual accuracy, and linguistic clarity across text, image, and video domains. Adept at crafting structured feedback and written assessments to enhance model understanding and real-world alignment. Strong foundation in data science, prompt design, and NLP, with a passion for contributing to frontier research that advances human-AI collaboration and model reliability.

Labeling Experience

Sentiment and Intent Classification for Conversational AI Systems

MindriftTextSegmentationClassification

Annotated conversational datasets to train intent-detection and sentiment-analysis models. Tasks involved classifying user utterances into predefined intent categories (e.g., request, complaint, inquiry) and rating tone as positive, neutral, or negative. Reviewed AI responses for empathy, coherence, and helpfulness. Maintained project throughput of over 500 samples per day with continuous quality monitoring. Adhered to project-specific accuracy benchmarks above 98% through double-blind review and feedback refinement cycles.

Annotated conversational datasets to train intent-detection and sentiment-analysis models. Tasks involved classifying user utterances into predefined intent categories (e.g., request, complaint, inquiry) and rating tone as positive, neutral, or negative. Reviewed AI responses for empathy, coherence, and helpfulness. Maintained project throughput of over 500 samples per day with continuous quality monitoring. Adhered to project-specific accuracy benchmarks above 98% through double-blind review and feedback refinement cycles.

2025

Text Reasoning and Quality Alignment for LLM Outputs

MercorTextText GenerationRLHF

Participated in large-scale evaluation of AI-generated text to assess reasoning accuracy, factual alignment, and linguistic clarity. Tasks included writing custom prompts, reviewing model completions, and assigning evaluation scores based on defined rubrics. Provided detailed rationales and structured feedback to improve model alignment and reasoning depth. The project covered thousands of text pairs and required high consistency, bias avoidance, and objective analysis. Quality was ensured through peer reviews and calibration sessions to maintain uniform standards across annotators.

Participated in large-scale evaluation of AI-generated text to assess reasoning accuracy, factual alignment, and linguistic clarity. Tasks included writing custom prompts, reviewing model completions, and assigning evaluation scores based on defined rubrics. Provided detailed rationales and structured feedback to improve model alignment and reasoning depth. The project covered thousands of text pairs and required high consistency, bias avoidance, and objective analysis. Quality was ensured through peer reviews and calibration sessions to maintain uniform standards across annotators.

2025 - 2025

Multimodal Content Evaluation (Image and Video Description Alignment)

Label StudioVideoBounding BoxPolygon

Evaluated AI-generated captions, short video descriptions, and motion-based visual reconstructions for accuracy, tone, and contextual alignment. Responsibilities included verifying object presence, assessing emotional realism in faces, and measuring quality metrics such as Visual Fidelity (VF), Motion and Dynamic Quality (MDQ), and Lip Sync Naturalness (LSQN). Delivered structured reports on image–text consistency and style alignment. The project spanned multiple annotation rounds with cross-validation to maintain over 95% accuracy across annotators.

Evaluated AI-generated captions, short video descriptions, and motion-based visual reconstructions for accuracy, tone, and contextual alignment. Responsibilities included verifying object presence, assessing emotional realism in faces, and measuring quality metrics such as Visual Fidelity (VF), Motion and Dynamic Quality (MDQ), and Lip Sync Naturalness (LSQN). Delivered structured reports on image–text consistency and style alignment. The project spanned multiple annotation rounds with cross-validation to maintain over 95% accuracy across annotators.

2024 - 2025

Education

Work History