Soma S.

Key Skills

Software

CVAT

Data Annotation Tech

Labelbox

LabelImg

Top Subject Matter

I am writing to introduce myself as a specialist in data labeling. My experience and skills in this field include [mention specific skills or areas of expertise, e.g., image annotation, natural language processing data tagging, quality assurance in labeling processes, etc

Top Data Types

Audio

Image

Text

Top Task Types

Bounding Box

Classification

Data Collection

Polygon

Polyline

Freelancer Overview

With over 6 years of dedicated experience in data labeling and AI training data, I bring a deep understanding of the intricacies involved in preparing high-quality datasets for machine learning models. My expertise encompasses a wide range of annotation tasks including image, video, and text data, utilizing advanced annotation tools such as Labelbox, Amazon SageMaker Ground Truth, and custom-built annotation platforms. I am proficient in developing annotation guidelines and workflows tailored to specific project requirements, ensuring consistency and accuracy in labeled datasets. Throughout my career, I have led and contributed to numerous complex projects across various domains including computer vision, natural language processing, and speech recognition. I have a proven track record of successfully managing large-scale annotation projects, collaborating closely with interdisciplinary teams to deliver annotated datasets that meet stringent quality standards and project timelines. Additionally, my skills in data preprocessing, quality assurance, and data validation have been instrumental in enhancing the efficacy and performance of machine learning models in real-world applications.

Labeling Experience

Data Labeling

CVATImageBounding Box

This project involves the annotation of 3,000 images for the purpose of improving object detection performance within a UK road environment. The annotated dataset will be used to train and evaluate a YOLOv11s (Ultralytics) model, with the goal of achieving over 90% mAP. Current annotations have resulted in approximately 85% mAP, indicating a need for higher-quality labeling standards—particularly more precise bounding box placement. Annotators must label only the following object categories: UK Vehicles Cars Buses Trucks Vans Each bounding box must meet high-precision criteria suitable for training a YOLOv11s model. Incorrect, loose, or misaligned annotations are likely contributors to the current performance cap (~85% mAP); therefore, improved precision is essential. Annotators should review each image thoroughly before submission to ensure quality

This project involves the annotation of 3,000 images for the purpose of improving object detection performance within a UK road environment. The annotated dataset will be used to train and evaluate a YOLOv11s (Ultralytics) model, with the goal of achieving over 90% mAP. Current annotations have resulted in approximately 85% mAP, indicating a need for higher-quality labeling standards—particularly more precise bounding box placement. Annotators must label only the following object categories: UK Vehicles Cars Buses Trucks Vans Each bounding box must meet high-precision criteria suitable for training a YOLOv11s model. Incorrect, loose, or misaligned annotations are likely contributors to the current performance cap (~85% mAP); therefore, improved precision is essential. Annotators should review each image thoroughly before submission to ensure quality

2025 - 2025

image annotation

LabelimgImageBounding Box

This project focuses on annotating invoices and bills of lading in the transportation sector, utilizing advanced data labeling techniques for car and human annotation. The scope includes accurately labeling key elements such as invoice numbers, dates, items shipped, and recipient details within documents. Additionally, the project involves annotating transport vehicles (cars, trucks, etc.) Accurately tagging and annotating invoice documents to identify and extract critical information crucial for financial and operational processes. The project manages a substantial volume of documents and multimedia data, requiring scalable annotation workflows to meet project timelines and deliverable Emphasis on quality assurance with rigorous validation processes to ensure annotation accuracy and consistency across datasets.

This project focuses on annotating invoices and bills of lading in the transportation sector, utilizing advanced data labeling techniques for car and human annotation. The scope includes accurately labeling key elements such as invoice numbers, dates, items shipped, and recipient details within documents. Additionally, the project involves annotating transport vehicles (cars, trucks, etc.) Accurately tagging and annotating invoice documents to identify and extract critical information crucial for financial and operational processes. The project manages a substantial volume of documents and multimedia data, requiring scalable annotation workflows to meet project timelines and deliverable Emphasis on quality assurance with rigorous validation processes to ensure annotation accuracy and consistency across datasets.

2022

Data Annotation/Data Labelling

LabelboxImageBounding BoxPolygon

Project Description: Data Annotation and data labelling and bounding box This project revolves around annotating diverse datasets for accurate recognition of cars, humans, and text elements using advanced annotation techniques including bounding boxes, keypoints, and text tags. The annotations are crucial for training and enhancing machine learning models that support applications in autonomous driving, surveillance, and document analysis. The project manages a substantial dataset comprising thousands of annotated images and hours of video footage, necessitating scalable annotation processes to meet project timelines and quality benchmarks. Quality assurance measures are implemented throughout, ensuring annotations adhere to defined standards and guidelines for accuracy and consistency. The project emphasizes delivering high-quality annotated datasets that enhance the performance and robustness of AI models in real-world applications

Project Description: Data Annotation and data labelling and bounding box This project revolves around annotating diverse datasets for accurate recognition of cars, humans, and text elements using advanced annotation techniques including bounding boxes, keypoints, and text tags. The annotations are crucial for training and enhancing machine learning models that support applications in autonomous driving, surveillance, and document analysis. The project manages a substantial dataset comprising thousands of annotated images and hours of video footage, necessitating scalable annotation processes to meet project timelines and quality benchmarks. Quality assurance measures are implemented throughout, ensuring annotations adhere to defined standards and guidelines for accuracy and consistency. The project emphasizes delivering high-quality annotated datasets that enhance the performance and robustness of AI models in real-world applications

2019 - 2022

Education

Work History