Nur Mustabsyirah Bin Binti Dzulkefli

Key Skills

Software

Clickworker

CVAT

Label Studio

OneForma

Top Subject Matter

No subject matter listed

Top Data Types

Image

Medical Dicom

Text

Top Label Types

Bounding Box

Entity Ner Classification

Polygon

Freelancer Overview

I have hands-on experience in AI training data through a combination of structured platform assessments and self-directed projects focused on data labeling and AI output evaluation. I have completed AI evaluation and annotation assessments with platforms such as Welocalize, and built independent projects where I reviewed AI-generated responses for accuracy, reasoning quality, clarity, instruction-following, and guideline compliance. My work involved identifying factual errors, logical gaps, ambiguity, bias, and safety concerns using industry-style rubrics similar to real-world AI vendor workflows. What sets me apart is my strong attention to detail, analytical thinking, and healthcare background, which trained me to evaluate information systematically, consider edge cases, and prioritize accuracy and safety. I am comfortable working with complex or ambiguous tasks, documenting clear justifications, and following detailed annotation guidelines. This combination of structured evaluation practice, self-initiative, and domain exposure allows me to contribute reliable, high-quality feedback to improve AI model performance.

Labeling Experience

POLYGON SEGMENTATION OF KNEE XRAY IMAGED

Label StudioMedical DicomPolygon

. Project Overview This project focuses on polygon-based segmentation of knee X-ray images to map key anatomical structures such as the femur, tibia, patella, and joint space. The goal is to produce high-quality, medically consistent annotations suitable for training and validating AI models in medical imaging, including osteoarthritis grading, landmark detection, and joint space measurements. 2. Objective ● To generate precise polygon segmentation masks of major knee structures. ● To build a clean, consistent annotated dataset usable for research or medical AI development. ● To follow anatomical standards to ensure accuracy and reproducibility. 3. Tools Used ● Label Studio – Polygon annotation (You can adjust the tools depending on what you actually used.) 4. Dataset Description ● Total Images: 30 knee X-rays ● Views: AP (Anteroposterior), Lateral (if included), Skyline View ● Format: PNG/JPG or converted DICOM ● Image Criteria: ○ Clear visualization of joint space

. Project Overview This project focuses on polygon-based segmentation of knee X-ray images to map key anatomical structures such as the femur, tibia, patella, and joint space. The goal is to produce high-quality, medically consistent annotations suitable for training and validating AI models in medical imaging, including osteoarthritis grading, landmark detection, and joint space measurements. 2. Objective ● To generate precise polygon segmentation masks of major knee structures. ● To build a clean, consistent annotated dataset usable for research or medical AI development. ● To follow anatomical standards to ensure accuracy and reproducibility. 3. Tools Used ● Label Studio – Polygon annotation (You can adjust the tools depending on what you actually used.) 4. Dataset Description ● Total Images: 30 knee X-rays ● Views: AP (Anteroposterior), Lateral (if included), Skyline View ● Format: PNG/JPG or converted DICOM ● Image Criteria: ○ Clear visualization of joint space

2025 - 2025

Health Article Named Entity Recognition

Label StudioTextEntity Ner Classification

Project Overview This project demonstrates a medical Named Entity Recognition (NER) workflow built from a scientific radiology article titled: “Imaging of the Posttreatment Head and Neck: Expected Findings and Potential Complications.” The goal is to extract clinically meaningful entities relevant to posttreatment imaging, complications, and head and neck anatomy. This project is designed to show competency in: ● Clinical NLP annotation ● Radiology entity identification ● Label Studio NER workflow ● Designing annotation schema and guidelines ● Creating dataset for AI/LLM training 🎯 2. Objective To annotate key medical concepts related to: ● TYPES OF DISEASE ● BODY PART ● IMAGING MODALITY ● TYPE OF TREATMENT ● IMAGE FINDING ● HEALTH CONDITION These entities are commonly used in: ● Clinical summarisation ● Radiology report structuring ● Medical chatbot training ● Information extraction tasks 🏷️ 3. Entity Categories Used Entity Label Description BODY_PART Specific anatomical structures

Project Overview This project demonstrates a medical Named Entity Recognition (NER) workflow built from a scientific radiology article titled: “Imaging of the Posttreatment Head and Neck: Expected Findings and Potential Complications.” The goal is to extract clinically meaningful entities relevant to posttreatment imaging, complications, and head and neck anatomy. This project is designed to show competency in: ● Clinical NLP annotation ● Radiology entity identification ● Label Studio NER workflow ● Designing annotation schema and guidelines ● Creating dataset for AI/LLM training 🎯 2. Objective To annotate key medical concepts related to: ● TYPES OF DISEASE ● BODY PART ● IMAGING MODALITY ● TYPE OF TREATMENT ● IMAGE FINDING ● HEALTH CONDITION These entities are commonly used in: ● Clinical summarisation ● Radiology report structuring ● Medical chatbot training ● Information extraction tasks 🏷️ 3. Entity Categories Used Entity Label Description BODY_PART Specific anatomical structures

2025 - 2025

HOUSEHOLD BOUNDING BOX PROJECT

CVATImageBounding Box

This project involves creating a mini Object Detection dataset using bounding box annotation on 52 household images. Objects labeled include keys, mugs, shoes, books, laptops, and mobile phones. Annotations were performed using CVAT (web version), and the final dataset was exported in COCO JSON format. This documentation includes objectives, dataset preparation, annotation workflow, annotation guidelines, COCO snippet samples, quality control checks, and project reflections.

This project involves creating a mini Object Detection dataset using bounding box annotation on 52 household images. Objects labeled include keys, mugs, shoes, books, laptops, and mobile phones. Annotations were performed using CVAT (web version), and the final dataset was exported in COCO JSON format. This documentation includes objectives, dataset preparation, annotation workflow, annotation guidelines, COCO snippet samples, quality control checks, and project reflections.

2025 - 2025

Education

Work History