Harshil S.

Key Skills

Software

CVAT

Roboflow

Labelbox

Label Studio

Top Subject Matter

Architectural CAD and Computer Vision

Satellite Imagery

Object Detection

Top Data Types

Document

Image

Video

Top Task Types

Polygon

Segmentation

Entity (NER) Classification

Bounding Box

Classification

Polyline

Object Detection

Fine-tuning

Text Generation

Text Summarization

Freelancer Overview

Data Annotation Specialist & AI Engineer. Brings 1+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include CVAT, QGIS, and Internal. Education includes Bachelor of Engineering, Hasmukh Goswami College of Engineering, Monark University (2021). AI-training focus includes data types such as Geospatial, Tiled Imagery, and Document and labeling workflows including Polygon, Segmentation, and Entity (NER) Classification.

Labeling Experience

Document Annotation & OCR Validation (AutoTenderAI Project)

DocumentEntity (NER) Classification

Validated and annotated large volumes of unstructured tender documents, correcting OCR outputs and extracting key eligibility criteria. Enhanced dataset quality for an LLM-based information retrieval pipeline focused on document intelligence. Performed entity recognition and qualification annotation for fine-tuning retrieval systems. • Labeled essential eligibility and compliance entities in scanned procurement documents. • Used internal or proprietary annotation tools for OCR validation and summary extraction. • Assisted in the data preparation workflow for a RAG AI application. • Enhanced the accuracy of document extraction pipelines through hands-on annotation work.

2025 - Present

Data Annotation Specialist & AI Engineer

CVATPolygon

Led complex data annotation and model training for architectural CAD pipelines, producing high-fidelity ground truth for computer vision tasks. Executed precise polygon segmentation and bounding box annotations for CAD elements, such as elevation drawings and symbols. Managed end-to-end dataset creation, focusing on area segmentations and symbol detection to drive robust detection models. • Carried out detailed CAD symbol detection and area drawing detection for over 25 classes. • Used CVAT and Label Studio to conduct polygon and bounding box labeling on architectural images and drawings. • Supported model training pipeline with meticulous annotation and dataset QA. • Collaborated with a team to maintain annotation guidelines and high label accuracy.

2025 - Present

AI Data Labeling & Research Intern

Segmentation

Created high-quality training datasets by manually annotating and segmenting complex road networks from aerial and satellite imagery. Conducted bounding box annotations for drone-based object detection and tracking tasks. Executed detailed text/OCR labeling for license plate detection systems to facilitate high-accuracy model training. • Led the annotation of geospatial imagery across five global cities, including Mumbai and Paris. • Used QGIS and Google Earth Engine for manual image segmentation and bounding box work. • Performed quality checks to ensure labeling precision and consistency. • Contributed directly to mapping and object detection models with reliable labels.

2024 - 2025

Education

Work History