Abhinav Tiwari

Key Skills

Software

Appen

Clickworker

CVAT

Google Cloud Vertex AI

Labelbox

Label Studio

OneForma

Scale AI

SuperAnnotate

Toloka

Internal/Proprietary Tooling

Top Subject Matter

Image Annotation for AI Training

Prompt and Response Evaluation for LLMs

Translation and Localization for AI Models

Top Data Types

Audio

Document

Image

Text

Video

Top Task Types

Action Recognition

Audio Recording

Bounding Box

Classification

Emotion Recognition

Entity Ner Classification

Evaluation Rating

Fine Tuning

Object Detection

Polygon

Polyline

Prompt Response Writing SFT

RLHF

Segmentation

Transcription

Translation Localization

Freelancer Overview

I am an AI Data Specialist with over five years of hands-on experience in data annotation, labeling, and quality assurance for speech, NLP, and multimodal AI projects. My expertise includes Hindi audio and AV recording, transcription QA, prompt and response writing for LLMs, and detailed multimodal annotation across text, audio, image, and video datasets. I have contributed to large-scale datasets powering speech recognition, conversational AI, and generative systems, ensuring data quality and linguistic accuracy through rigorous QA processes. I am proficient with tools like Labelbox, Label Studio, and Audacity, and have experience in RLHF-style prompt evaluation, localization, and multilingual model fine-tuning. My practical approach and deep understanding of AI training workflows enable me to deliver high-quality, culturally relevant data that drives real-world AI applications.

Labeling Experience

Data Annotation

CVATImageBounding BoxPolygon

In this project, I was tasked with annotating a large dataset of images to train a machine learning model for Image Classification. Using the CVAT labeling tool, I categorized images into specific classes such as vehicles, buildings, animals, and natural landscapes. The dataset was diverse, consisting of images from urban, rural, and natural environments, each requiring accurate labeling to train the model to recognize and classify objects in various contexts.

In this project, I was tasked with annotating a large dataset of images to train a machine learning model for Image Classification. Using the CVAT labeling tool, I categorized images into specific classes such as vehicles, buildings, animals, and natural landscapes. The dataset was diverse, consisting of images from urban, rural, and natural environments, each requiring accurate labeling to train the model to recognize and classify objects in various contexts.

2024

Data Annotation

CVATVideoClassification

In this project, I was responsible for annotating and labeling video data to assist in training machine learning models for autonomous vehicle systems. Using the CVAT labeling tool, I applied various types of annotations, including Bounding Box for vehicle detection, Polyline for road boundaries, Segmentation for identifying lanes and road surfaces, and Action Recognition for labeling vehicle movements such as turning, stopping, or lane changing.

In this project, I was responsible for annotating and labeling video data to assist in training machine learning models for autonomous vehicle systems. Using the CVAT labeling tool, I applied various types of annotations, including Bounding Box for vehicle detection, Polyline for road boundaries, Segmentation for identifying lanes and road surfaces, and Action Recognition for labeling vehicle movements such as turning, stopping, or lane changing.

2024

AI Evaluator

LabelboxTextEntity Ner Classification

collaborated with Alignerr on the Hindi language labelbox.

collaborated with Alignerr on the Hindi language labelbox.

2024

AI Trainer

LabelboxTextTranslation LocalizationPrompt Response Writing SFT

As an AI Trainer, I worked on a project focused on improving natural language processing models by creating and refining text prompts and responses in Hindi. The tasks involved translating English prompts into Hindi and evaluating LLM-generated responses to ensure they were accurate, contextually relevant, and linguistically appropriate. I used Labelbox to manage and organize the data labeling tasks, ensuring consistency across large datasets and maintaining high standards for accuracy and cultural sensitivity. The project required constant collaboration with quality control teams to ensure the AI model’s improvements aligned with client specifications.

As an AI Trainer, I worked on a project focused on improving natural language processing models by creating and refining text prompts and responses in Hindi. The tasks involved translating English prompts into Hindi and evaluating LLM-generated responses to ensure they were accurate, contextually relevant, and linguistically appropriate. I used Labelbox to manage and organize the data labeling tasks, ensuring consistency across large datasets and maintaining high standards for accuracy and cultural sensitivity. The project required constant collaboration with quality control teams to ensure the AI model’s improvements aligned with client specifications.

2024

Crowdgen

AppenTextTranslation LocalizationPrompt Response Writing SFT

At Crowdgen, I worked on refining and enhancing large language models (LLMs) by generating and evaluating text responses across various domains. My tasks included writing prompts, reviewing model outputs, and translating English content into Hindi to ensure linguistic and contextual accuracy. I also evaluated the coherence and relevance of LLM-generated responses, contributing to the continuous improvement of the model’s language comprehension and output quality. Using Appen's platform, I helped manage a high volume of tasks, ensuring that all data met strict quality control standards for further model training.

At Crowdgen, I worked on refining and enhancing large language models (LLMs) by generating and evaluating text responses across various domains. My tasks included writing prompts, reviewing model outputs, and translating English content into Hindi to ensure linguistic and contextual accuracy. I also evaluated the coherence and relevance of LLM-generated responses, contributing to the continuous improvement of the model’s language comprehension and output quality. Using Appen's platform, I helped manage a high volume of tasks, ensuring that all data met strict quality control standards for further model training.

2023

Show more

Education

Work History