Umaid Zakaria

Key Skills

Software

CVAT

Labelbox

Label Studio

Doccano

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Bounding Box

Entity Ner Classification

Evaluation Rating

Freelancer Overview

I am an AI Training & Data Labeling Specialist with hands-on experience in natural language processing (NLP), large language models (LLMs), and computer vision projects. At Invisible Technologies, I worked as an Advanced AI Trainer where I contributed to the training and fine-tuning of generative AI systems, specializing in prompt engineering, evaluation, error mitigation, and RLHF-based annotation. My expertise also extends to text summarization, entity recognition (NER), machine translation, and cultural sensitivity in AI systems, ensuring models perform accurately and responsibly. In addition to NLP, I have contributed to computer vision labeling workflows, creating and refining datasets for object detection, bounding box annotation, segmentation, and video tracking using tools like CVAT, Label Studio, and Labelbox. My experience bridges both text and image/video modalities, giving me a versatile foundation to contribute to AI projects across industries including healthcare imaging, autonomous systems, and generative AI evaluation.

Labeling Experience

NLP Dataset Annotation – Text Classification & NER

DoccanoTextEntity Ner ClassificationQuestion Answering

Created labeled corpora for entity recognition, summarization, and QA tasks. Generated token-level annotations to support NLP models in classification, summarization, and text comprehension. Produced high-quality annotations following strict guidelines to ensure data consistency and reliability.

Created labeled corpora for entity recognition, summarization, and QA tasks. Generated token-level annotations to support NLP models in classification, summarization, and text comprehension. Produced high-quality annotations following strict guidelines to ensure data consistency and reliability.

2024 - 2025

Video Action Recognition & Tracking Annotations

CVATVideoBounding BoxPoint Key Point

Labeled video datasets by creating tubelets for consistent object tracking across frames. Annotated bounding boxes and key points to capture movement and activity recognition. Focused on temporal consistency to enable robust action recognition and trajectory prediction models.

Labeled video datasets by creating tubelets for consistent object tracking across frames. Annotated bounding boxes and key points to capture movement and activity recognition. Focused on temporal consistency to enable robust action recognition and trajectory prediction models.

2024 - 2024

Object Detection Dataset Annotation (YOLO & R-CNN)

CVATImageBounding BoxSegmentation

Annotated and preprocessed large-scale image datasets in YOLO format for object detection tasks. Created accurate bounding boxes and segmentation masks for multiple object categories. Enhanced dataset robustness with augmentation techniques, ensuring consistency across annotations. The labeled data powered a CNN/R-CNN pipeline that achieved 99% model accuracy.

Annotated and preprocessed large-scale image datasets in YOLO format for object detection tasks. Created accurate bounding boxes and segmentation masks for multiple object categories. Enhanced dataset robustness with augmentation techniques, ensuring consistency across annotations. The labeled data powered a CNN/R-CNN pipeline that achieved 99% model accuracy.

2024 - 2024

Education

Work History