Tricia Marteney

Key Skills

Software

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Image

Top Label Types

Text Generation

Freelancer Overview

I am a highly analytical QA Reviewer and Image Data Validation Specialist with over 10 years of experience supporting AI training data quality for leading technology companies. My background includes auditing and correcting image-based question answering datasets, identifying visual-label mismatches, and ensuring precise extraction and verification of text from images. I have worked extensively with annotation platforms like Labelbox, CVAT, and Scale Studio, and am skilled in using bounding box validation and OCR review tools for computer vision and generative AI projects. My expertise spans data annotation, error detection, standalone answer rewriting, and structured QA workflows, with a strong emphasis on linguistic precision and visual-text consistency. I consistently meet production targets while exceeding quality benchmarks in remote environments, and have published research on scalable quality assurance frameworks for multimodal AI systems.

Labeling Experience

Multimodal Image-Text QA Validation & Annotation for Enterprise AI Systems

CVATImageText Generation

Engaged in massive multimodal data labeling and validation for enterprise-level AI training pipelines related to image-based question answering and visual understanding. Tasks included validating image-text pairs, confirming the accuracy of extracted text, correcting incorrectly labeled objects and scenes, and validating visual-text consistency. Performed classification, answer rewriting, and quality assessment within tight annotation requirements. Identified ambiguous and unanswerable questions with rationale and used structured assessment frameworks to ensure dataset quality. Monitored more than 30,000 items while meeting production requirements of 7-8 minutes per item and 98% quality adherence.

Engaged in massive multimodal data labeling and validation for enterprise-level AI training pipelines related to image-based question answering and visual understanding. Tasks included validating image-text pairs, confirming the accuracy of extracted text, correcting incorrectly labeled objects and scenes, and validating visual-text consistency. Performed classification, answer rewriting, and quality assessment within tight annotation requirements. Identified ambiguous and unanswerable questions with rationale and used structured assessment frameworks to ensure dataset quality. Monitored more than 30,000 items while meeting production requirements of 7-8 minutes per item and 98% quality adherence.

2019 - 2024

Education

Work History