Clinton O.

Key Skills

Software

Label Studio

Other

Prodigy

CVAT

iMerit

Top Subject Matter

AI - Multimodal Data Annotation

AI - Content Quality Analysis

English - Writing

Top Data Types

Text

Audio

Image

Video

Document

Top Task Types

Transcription

Entity (NER) Classification

Bounding Box

Polygon

Segmentation

Classification

Object Detection

Polyline

Evaluation/Rating

RLHF

Text Generation

Text Summarization

Fine-tuning

Tracking

Audio Recording

Freelancer Overview

Detail-oriented AI Data Trainer and Annotation Specialist with hands-on experience across text, audio, image, and video datasets. Proven ability to deliver high-quality labeled data for AI model training, evaluation, and fine-tuning. Experienced in object detection, object tracking, NLP annotation, audio transcription, and RLHF workflows. Known for maintaining 98%+ accuracy in image and text annotation and 90%+ tracking accuracy in video datasets in production environments.

Labeling Experience

Data Annotation Specialist (Catalog Matching Project)

OtherDocumentClassification

For Blend’s Catalog Matching Project, I annotated and matched over 5,000 data entries by comparing page and catalog names using domain knowledge and metadata analysis. Binary classification decisions required strict adherence to project guidelines and detailed reasoning. I played a vital role in ensuring dataset integrity for catalog management AI. • Identified edge cases including naming mismatches and ambiguities. • Maintained high accuracy and consistency across a large document set. • Contributed to ongoing validation and quality refinement processes. • Enhanced metadata matching for improved dataset reliability.

2026 - Present

AI Image Quality Evaluator (NiDRA WF11 Project)

iMeritImage

As an AI Image Quality Evaluator with iMerit on the NiDRA WF11 Project, I assessed AI-generated image quality using structured metrics. Tasks focused on identifying defects and comparing images for perceptual quality. My work ensured consistent visual assessment standards and actionable feedback incorporation. • Evaluated images for noise, sharpness, exposure, and visual artifacts. • Used structured IQA principles for objective scoring and consistency. • Maintained over 95% accuracy in quality evaluation tasks. • Processed high volumes efficiently while upholding standards.

2026 - Present

Data Annotation Specialist

CVATImageEntity (NER) ClassificationRLHF

Annotated multimodal datasets (image, video, text, audio) for varied AI training and evaluation projects. Executed tasks including bounding box, polygon segmentation, classification, and object tracking for thousands of images and video clips. Ensured high annotation accuracy and adapted to diverse guidelines and domains. • Used CVAT and Label Studio extensively for annotation work. • Worked on traffic, robotics, retail, and general object detection tasks. • Achieved 98%+ accuracy in images/text and 90%+ in video tracking. • Demonstrated rapid learning of new tools and project requirements.

2024 - Present

AI Data Trainer & Content Evaluator

Label StudioText

In this role, I evaluated AI-generated text responses to assess clarity, factual accuracy, and tone. I created and tested text prompts to gauge model reasoning and overall creativity. I annotated datasets for semantic meaning, intent, and sentiment, ensuring quality standards were met. • Performed manual QA to maintain a 98% accuracy benchmark. • Refined conversational flow using UX insights. • Processed over 500 AI prompts for evaluation. • Reported findings for performance improvements.

2024 - Present

Traffic Flow and Pedestrian Tracking for Autonomous Navigation

VideoBounding Box

I worked on a video annotation project aimed at improving how an autonomous driving system understands movement and behavior over time. The focus was on helping the model handle dynamic urban traffic, things like sudden lane changes, pedestrian crossings, and interactions at intersections. The dataset consisted of about 2,500 video clips (each 20–40 seconds long), extracted from dashcam footage across busy city routes. Unlike image tasks, this required frame-by-frame object tracking. We annotated and tracked vehicles, pedestrians, cyclists, and traffic lights across sequences, ensuring each object maintained a consistent ID throughout its movement. A big part of the work involved temporal consistency, for example, making sure a pedestrian entering from one frame was tracked accurately even if briefly occluded or partially out of view. We also labeled events, such as jaywalking, sudden stops, or turns, based on predefined criteria. Quality standards were strict. The project aimed for 90–93% tracking accuracy, measured by ID consistency and bounding box precision across frames. There was a two-stage QA process: initial peer review, followed by random audits from a senior team. Edge cases like heavy occlusion, motion blur, or night scenes were flagged and reviewed separately to maintain consistency with the guidelines.

2026 - 2026

Show more

Education

Work History