Abraham Aisosa

Key Skills

Software

Appen

Remotasks

Labelbox

Roboflow

Label Studio

CVAT

Top Subject Matter

Computer Vision

Autonomous Vehicles

Natural Language Processing

Top Data Types

Video

Image

Text

Document

Top Task Types

Object Detection

Classification

RLHF

Segmentation

Data Collection

Text Summarization

Bounding Box

Freelancer Overview

Data Labeller & Video Annotator. Brings 3+ years of professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Autonomous vehicles, computer vision, natural language processing, e-commerce, medical imaging, content safety & moderation. Education includes Bachelor of Engineering, University of Benin (2021). AI-training focus includes data types such as Video, Image, and Text and labeling workflows including Object Detection, Classification, and RLHF.

Labeling Experience

Data Labeller & Video Annotator

VideoObject Detection

Conducted frame-level video annotation and multi-object tracking using Atlas Capture for 500+ hours of footage. Maintained 97%+ accuracy on multimodal annotation tasks spanning video, image, and audio data for computer vision and autonomous systems. Refined annotation guidelines in collaboration with ML teams to improve training data quality and reduced inter-annotator disagreement by 30%. • Applied bounding boxes, semantic segmentation masks, polygon annotation, and activity classification labels across datasets. • Identified and escalated complex edge cases to enhance annotation consistency and model precision. • Quality-reviewed annotations with sustained sub-1% error rates and contributed feedback to improve labeling rubrics. • Supported training pipelines for computer vision models in production environments.

Conducted frame-level video annotation and multi-object tracking using Atlas Capture for 500+ hours of footage. Maintained 97%+ accuracy on multimodal annotation tasks spanning video, image, and audio data for computer vision and autonomous systems. Refined annotation guidelines in collaboration with ML teams to improve training data quality and reduced inter-annotator disagreement by 30%. • Applied bounding boxes, semantic segmentation masks, polygon annotation, and activity classification labels across datasets. • Identified and escalated complex edge cases to enhance annotation consistency and model precision. • Quality-reviewed annotations with sustained sub-1% error rates and contributed feedback to improve labeling rubrics. • Supported training pipelines for computer vision models in production environments.

2025 - Present

Data Collection & Labelling Specialist (Contract)

AppenImageClassification

Curated and labeled speech, text, and image datasets for global AI training projects at Appen. Delivered over 150,000 high-quality labeled data points for tasks such as search relevance, sentiment analysis, and named entity recognition. Reviewed peer annotations during QA cycles and refined annotation rubrics to reduce rework rates by 25%. • Labeled multimodal data including text, speech, and images to support AI model development for enterprise clients. • Ensured dataset quality above platform thresholds, contributing to client AI improvements. • Provided actionable annotation feedback and validated complex cases for dataset consistency. • Maintained throughput and accuracy in high-volume labeling environments.

Curated and labeled speech, text, and image datasets for global AI training projects at Appen. Delivered over 150,000 high-quality labeled data points for tasks such as search relevance, sentiment analysis, and named entity recognition. Reviewed peer annotations during QA cycles and refined annotation rubrics to reduce rework rates by 25%. • Labeled multimodal data including text, speech, and images to support AI model development for enterprise clients. • Ensured dataset quality above platform thresholds, contributing to client AI improvements. • Provided actionable annotation feedback and validated complex cases for dataset consistency. • Maintained throughput and accuracy in high-volume labeling environments.

2024 - Present

E-Commerce Product Image Classification Contributor

AppenImageClassification

Labeled 50,000+ e-commerce product images for category and attribute extraction to power AI recommendation systems. Maintained 98% label accuracy on a stratified QA sample, supporting a 12% click-through rate lift post-deployment. Tagged product images with multi-class attributes including color, material, style, and condition. • Performed large-scale image annotation for retail recommendation engines. • Sustained high-quality standards and accurate attribute tagging in diverse product sets. • Validated annotation batches for integrity before model training phases. • Contributed to ML-driven business outcomes through scalable image labeling.

Labeled 50,000+ e-commerce product images for category and attribute extraction to power AI recommendation systems. Maintained 98% label accuracy on a stratified QA sample, supporting a 12% click-through rate lift post-deployment. Tagged product images with multi-class attributes including color, material, style, and condition. • Performed large-scale image annotation for retail recommendation engines. • Sustained high-quality standards and accurate attribute tagging in diverse product sets. • Validated annotation batches for integrity before model training phases. • Contributed to ML-driven business outcomes through scalable image labeling.

2023 - 2024

LLM Safety & Alignment Dataset Annotator

RemotasksTextRLHF

Ranked and scored more than 1,500 AI-generated response pairs for LLM safety and alignment datasets in RLHF tasks. Judged helpfulness, harmlessness, and honesty, directly contributing to reward model training for large language models. Maintained 99%+ precision in flagging unsafe or non-compliant outputs to improve AI safety. • Conducted granular safety, accuracy, and compliance reviews of language model generations. • Identified and flagged harmful or instruction-breaking content. • Supported training and evaluation of language model reward functions. • Maintained high accuracy in RLHF response ranking for LLM model alignment.

Ranked and scored more than 1,500 AI-generated response pairs for LLM safety and alignment datasets in RLHF tasks. Judged helpfulness, harmlessness, and honesty, directly contributing to reward model training for large language models. Maintained 99%+ precision in flagging unsafe or non-compliant outputs to improve AI safety. • Conducted granular safety, accuracy, and compliance reviews of language model generations. • Identified and flagged harmful or instruction-breaking content. • Supported training and evaluation of language model reward functions. • Maintained high accuracy in RLHF response ranking for LLM model alignment.

2023 - 2024

Autonomous Vehicle Scene Understanding Project Contributor

VideoSegmentation

Annotated over 300 hours of dashcam video for autonomous vehicle scene understanding projects. Labeled lane markings, pedestrian bounding boxes, traffic sign classes, and vehicle tracking IDs with a focus on temporal consistency. Achieved agreement above 96% in quality assurance reviews for object and vehicle tracking. • Applied semantic segmentation, object tracking, and classification across frame sequences. • Supported annotation of real-world driving environments in challenging conditions. • Enhanced dataset quality for autonomous vehicle perception model training. • Contributed to reduced error rates in vehicle tracking and scene labeling QC.

Annotated over 300 hours of dashcam video for autonomous vehicle scene understanding projects. Labeled lane markings, pedestrian bounding boxes, traffic sign classes, and vehicle tracking IDs with a focus on temporal consistency. Achieved agreement above 96% in quality assurance reviews for object and vehicle tracking. • Applied semantic segmentation, object tracking, and classification across frame sequences. • Supported annotation of real-world driving environments in challenging conditions. • Enhanced dataset quality for autonomous vehicle perception model training. • Contributed to reduced error rates in vehicle tracking and scene labeling QC.

2023 - 2024

Show more

Education

Work History