Harsh Yadav

Key Skills

Software

Labelbox

Mindrift

Other

Top Subject Matter

Large Language Model Safety

Reasoning Domain Expertise

and Creative Writing

Top Data Types

Audio

Video

Text

Top Task Types

Text Generation

Translation/Localization

Red Teaming

RLHF

Transcription

Freelancer Overview

Remote Freelance AI Data Annotator & Model Evaluation Specialist. Brings 1+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Mindrift, Labelbox, and Other. Education includes Bachelor of Technology, Dr. Shakuntala Misra National Rehabilitation University (2025). AI-training focus includes data types such as Text and Audio and labeling workflows including Red Teaming, RLHF, and Transcription.

Labeling Experience

Remote Freelance AI Data Annotator & Model Evaluation Specialist

MindriftTextRed Teaming

As a Remote Freelance AI Data Annotator & Model Evaluation Specialist at Mindrift, I specialize in red teaming, data labeling, and benchmarking of Large Language Models. I design and label datasets to enhance model reasoning and safety in logic-oriented and creative contexts. My work includes detailed adversarial testing and the creation of high-quality instruction datasets for model evaluation and improvement. • Conduct adversarial tests to identify LLM vulnerabilities and bias issues. • Curate and annotate diverse instruction datasets for enhanced reasoning. • Execute side-by-side model comparison for 3H benchmarks (helpfulness, honesty, harmlessness). • Enhance data quality and diversity for model fine-tuning and evaluation.

As a Remote Freelance AI Data Annotator & Model Evaluation Specialist at Mindrift, I specialize in red teaming, data labeling, and benchmarking of Large Language Models. I design and label datasets to enhance model reasoning and safety in logic-oriented and creative contexts. My work includes detailed adversarial testing and the creation of high-quality instruction datasets for model evaluation and improvement. • Conduct adversarial tests to identify LLM vulnerabilities and bias issues. • Curate and annotate diverse instruction datasets for enhanced reasoning. • Execute side-by-side model comparison for 3H benchmarks (helpfulness, honesty, harmlessness). • Enhance data quality and diversity for model fine-tuning and evaluation.

2026 - Present

AI Data Labeling & LLM Evaluation Project

LabelboxVideoText GenerationTranslation Localization

Worked on multilingual AI training tasks involving text, audio, and video data. Responsibilities included labeling, classifying, and moderating content for safety, accuracy, and policy compliance. Performed LLM evaluation, prompt testing, and quality rating in both Hindi and English. Ensured high-quality annotations with strict adherence to project guidelines and QA standards. Contributed to improving AI model performance, reducing hallucinations, and refining safety alignment through consistent, accurate annotations.

Worked on multilingual AI training tasks involving text, audio, and video data. Responsibilities included labeling, classifying, and moderating content for safety, accuracy, and policy compliance. Performed LLM evaluation, prompt testing, and quality rating in both Hindi and English. Ensured high-quality annotations with strict adherence to project guidelines and QA standards. Contributed to improving AI model performance, reducing hallucinations, and refining safety alignment through consistent, accurate annotations.

2025 - Present

Remote Transcription & Audio Data Specialist

OtherAudioTranscription

As a Freelance Remote Transcription & Audio Data Specialist, I produce precise 'ground truth' transcriptions to support Speech-to-Text model development. I structure and annotate audio datasets for high fidelity in machine learning tasks. My responsibilities include timestamping, speaker diarization, and acoustic tagging within audio data. • Generate accurate verbatim transcriptions for STT model training. • Apply detailed timestamping and speaker identification in datasets. • Execute acoustic tagging for nuanced audio event detection. • Format and quality-check audio data for ML readiness.

As a Freelance Remote Transcription & Audio Data Specialist, I produce precise 'ground truth' transcriptions to support Speech-to-Text model development. I structure and annotate audio datasets for high fidelity in machine learning tasks. My responsibilities include timestamping, speaker diarization, and acoustic tagging within audio data. • Generate accurate verbatim transcriptions for STT model training. • Apply detailed timestamping and speaker identification in datasets. • Execute acoustic tagging for nuanced audio event detection. • Format and quality-check audio data for ML readiness.

2024 - Present

Education

Work History