Manjeet Yadav

Key Skills

Software

Other

Google Cloud Vertex AI

CVAT

Labelbox

Roboflow

Snorkel AI

Top Subject Matter

Autonomous Vehicles

Computer Vision

Customer Reviews

Top Data Types

Image

Video

Text

Document

Top Task Types

Bounding Box

Text Generation

Evaluation/Rating

Classification

Freelancer Overview

Autonomous Vehicle Image Annotation. Core strengths include CVAT, Labelbox, and Roboflow. Education includes Bachelor of Computer Applications, IGNOU - Indira Gandhi National Open University (2026). AI-training focus includes data types such as Image and Text and labeling workflows including Bounding Box and Classification.

Labeling Experience

Data Annotation & Labeling Practice (Self Projects)

ImageClassification

Labeled and annotated image datasets as part of self-directed projects focused on AI training tasks. Ensured correct categorization and tagging of images to facilitate efficient machine learning workflows. Verified dataset quality for downstream model training and evaluation. • Assigned descriptive labels to image data for classification tasks • Inspected visual data for accuracy, errors, and consistency • Maintained annotation logs and records for project tracking • Enhanced datasets for use in AI and computer vision projects

Labeled and annotated image datasets as part of self-directed projects focused on AI training tasks. Ensured correct categorization and tagging of images to facilitate efficient machine learning workflows. Verified dataset quality for downstream model training and evaluation. • Assigned descriptive labels to image data for classification tasks • Inspected visual data for accuracy, errors, and consistency • Maintained annotation logs and records for project tracking • Enhanced datasets for use in AI and computer vision projects

2023 - Present

Data Annotation & Labeling Practice (Self Projects)

TextClassification

Annotated and labeled text datasets for AI model training using classification and error correction methodologies. Maintained high accuracy while reviewing and tagging large volumes of content according to project guidelines. Focused on improving model performance and data cleanliness through careful labeling work. • Performed data tagging and error identification on text data • Applied classification labels to diverse AI training datasets • Adhered strictly to annotation guidelines and quality standards • Worked independently with strong attention to detail

Annotated and labeled text datasets for AI model training using classification and error correction methodologies. Maintained high accuracy while reviewing and tagging large volumes of content according to project guidelines. Focused on improving model performance and data cleanliness through careful labeling work. • Performed data tagging and error identification on text data • Applied classification labels to diverse AI training datasets • Adhered strictly to annotation guidelines and quality standards • Worked independently with strong attention to detail

2023 - Present

Text Annotation for Sentiment Analysis

Snorkel AITextClassification

I manually annotated 120,000 customer review texts with multi-class sentiment and emotion labels for a sentiment analysis project. To accelerate throughput and improve consistency, I applied Snorkel weak supervision functions to automate label generation for a majority of the corpus. My efforts included refining an annotation guide and fine-tuning a BERT model for sentiment classification. • Reduced inter-annotator disagreement significantly by implementing a detailed annotation style guide. • Applied automated labeling to 60 percent of the total corpus using Snorkel. • Used CVAT and Snorkel AI tools for annotation and weak supervision labeling processes. • Fine-tuned a BERT base model, achieving strong F1 scores versus traditional classification methods.

I manually annotated 120,000 customer review texts with multi-class sentiment and emotion labels for a sentiment analysis project. To accelerate throughput and improve consistency, I applied Snorkel weak supervision functions to automate label generation for a majority of the corpus. My efforts included refining an annotation guide and fine-tuning a BERT model for sentiment classification. • Reduced inter-annotator disagreement significantly by implementing a detailed annotation style guide. • Applied automated labeling to 60 percent of the total corpus using Snorkel. • Used CVAT and Snorkel AI tools for annotation and weak supervision labeling processes. • Fine-tuned a BERT base model, achieving strong F1 scores versus traditional classification methods.

2026 - 2026

Autonomous Vehicle Image Annotation

CVATImageBounding Box

I annotated 85,000 dash cam and LiDAR images for autonomous driving research, ensuring adherence to ADAS standards with high accuracy. This project required consistent multi-pass quality assurance and conversion to standard formats for model training. My work improved data efficiency and increased model mean average precision for object detection tasks. • Labeled images using bounding boxes and polygons for 12 distinct classes such as vehicles, pedestrians, and cyclists. • Used CVAT, Labelbox, Roboflow, and custom pipelines to produce datasets in COCO JSON and YOLO formats. • Enabled reproducible pipelines for YOLO v8 fine-tuning by documenting the full annotation workflow. • Delivered less than 2 percent annotation error rate by following strict quality control protocols.

I annotated 85,000 dash cam and LiDAR images for autonomous driving research, ensuring adherence to ADAS standards with high accuracy. This project required consistent multi-pass quality assurance and conversion to standard formats for model training. My work improved data efficiency and increased model mean average precision for object detection tasks. • Labeled images using bounding boxes and polygons for 12 distinct classes such as vehicles, pedestrians, and cyclists. • Used CVAT, Labelbox, Roboflow, and custom pipelines to produce datasets in COCO JSON and YOLO formats. • Enabled reproducible pipelines for YOLO v8 fine-tuning by documenting the full annotation workflow. • Delivered less than 2 percent annotation error rate by following strict quality control protocols.

2025 - 2025

Education

Work History