Abdallah Ammar

Key Skills

Software

Appen

CrowdFlower

CVAT

Data Annotation Tech

Labelbox

OneForma

OpenCV AI Kit (OAK)

Prodigy

Roboflow

Scale AI

Toloka

Telus

Surge AI

Label Studio

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Document

Text

Top Task Types

Computer Programming Coding

Evaluation Rating

RLHF

Text Generation

Translation Localization

Freelancer Overview

AI Data Specialist with a PhD in Computer Science from Stanford University and extensive experience in data labeling, AI model training, and quality evaluation. I've annotated and optimized diverse datasets (text, image, audio) for NLP, speech recognition, and search algorithms at companies like DataTech Solutions and BrightMind AI. My expertise includes designing scalable labeling workflows, implementing rigorous QA checks, and collaborating with ML teams to enhance model performance—ensuring data quality aligns with human values. What sets me apart is my research-backed approach to reducing labeling errors by 30% through improved guidelines, hands-on experience with end-to-end data pipelines, and certifications in machine learning (Google, Microsoft). My work directly contributes to building safer, more accurate AI systems—perfectly aligning with Surge AI's mission.

Labeling Experience

Radiology Image Segmentation for Diagnostic AI

Scale AIImageSegmentationCuboid

Annotated 15,000+ CT/MRI scans to train a cancer detection model. Tasks included: Precise tumor segmentation with radiologist verification Classifying malignancy likelihood (Benign/Suspicious/Malignant) Measuring lesion dimensions via cuboid annotation Achieved 98% concordance with expert radiologists on test set

Annotated 15,000+ CT/MRI scans to train a cancer detection model. Tasks included: Precise tumor segmentation with radiologist verification Classifying malignancy likelihood (Benign/Suspicious/Malignant) Measuring lesion dimensions via cuboid annotation Achieved 98% concordance with expert radiologists on test set

2024 - 2024

Street Scene Annotation for Self-Driving AI

CVATImageBounding BoxPolygon

Led a team labeling 100,000+ street view images for an autonomous vehicle startup. Tasks included: Drawing precise bounding boxes around vehicles/pedestrians Semantic segmentation of road surfaces and lanes Quality control through consensus scoring (3 raters per image) Developed guidelines for edge cases (occluded objects, poor lighting)

Led a team labeling 100,000+ street view images for an autonomous vehicle startup. Tasks included: Drawing precise bounding boxes around vehicles/pedestrians Semantic segmentation of road surfaces and lanes Quality control through consensus scoring (3 raters per image) Developed guidelines for edge cases (occluded objects, poor lighting)

2023 - 2023

Multimodal NLP Dataset Annotation for LLM Fine-Tuning

Scale AIAudioEntity Ner ClassificationQuestion Answering

Led annotation of 50,000+ text and audio samples (patient queries/doctor responses) to fine-tune a healthcare LLM. Tasks included: Labeling medical entities (symptoms, medications) with 95% inter-rater agreement. Classifying intent (e.g., "diagnosis request" vs. "treatment advice") and emotion tones (urgency, concern). Writing synthetic prompt-response pairs to fill data gaps. Implementing QA checks: 10% random reviews by senior raters + automated consistency validation (Python scripts).

Led annotation of 50,000+ text and audio samples (patient queries/doctor responses) to fine-tune a healthcare LLM. Tasks included: Labeling medical entities (symptoms, medications) with 95% inter-rater agreement. Classifying intent (e.g., "diagnosis request" vs. "treatment advice") and emotion tones (urgency, concern). Writing synthetic prompt-response pairs to fill data gaps. Implementing QA checks: 10% random reviews by senior raters + automated consistency validation (Python scripts).

2023 - 2023

Search Relevance Rating for AI-Powered Engine

Surge AITextClassificationEvaluation Rating

Rated 20,000+ search query-result pairs for a major tech client, achieving 98% alignment with gold-standard benchmarks. Developed ambiguity-resolution guidelines adopted company-wide.

Rated 20,000+ search query-result pairs for a major tech client, achieving 98% alignment with gold-standard benchmarks. Developed ambiguity-resolution guidelines adopted company-wide.

2022 - 2023

Image Annotation

Label StudioImage

Used Label Studio for 6 months to annotate image data for OCR tasks, including drawing precise bounding boxes and polygons for text extraction.

Used Label Studio for 6 months to annotate image data for OCR tasks, including drawing precise bounding boxes and polygons for text extraction.

2022 - 2022

Show more

Education

Work History