Yamraj Puri

Key Skills

Software

Scale AI

Labelbox

Remotasks

Mercor

Mindrift

Toloka

Top Subject Matter

Large Language Models (LLMs)

AI Model Evaluation

Data Annotation

Top Data Types

Text

Image

Audio

Top Task Types

RLHF

Red Teaming

Transcription

Text Generation

Freelancer Overview

Data Labeling and Annotation Specialist with over 2+ years of experience helping teams train and evaluate AI models. I have worked on platforms such as Scale AI, Labelbox, and Remotasks. I have also served as a reviewer in Labelbox to perform second‑pass quality checks and resolve edge cases. In addition, I’ve labeled and evaluated AI voice assistants, reviewing conversational turns, checking transcription accuracy, and rating how well the assistant follows user instructions and safety policies. I am used to following detailed guidelines, documenting tricky examples, and consistently meeting both quality and productivity targets. I have a Master of Computer Applications (MCA) and a background in data engineering and automation, which helps me understand how labeled data powers downstream ML systems. I regularly use Python and SQL for light data analysis and quality checks, and I’m comfortable working remotely, handling sensitive data under NDA, and providing structured feedback that improves project rubrics and model behavior.

Labeling Experience

Data Labeling/Annotation Specialist

Scale AIText

This experience involved annotation and evaluation of AI-generated outputs for quality and accuracy. I worked with tools such as Scale AI, Labelbox, and Remotasks to assess text responses and perform data labeling tasks. The focus was on enhancing model alignment and refining large language model interactions for enterprise applications. • Conducted comprehensive evaluation and annotation of AI text outputs. • Utilized multiple data labeling platforms including Scale AI, Labelbox, and Remotasks. • Maintained high standards for data accuracy in model training workflows. • Collaborated with the team to optimize prompt engineering and refine AI responses.

This experience involved annotation and evaluation of AI-generated outputs for quality and accuracy. I worked with tools such as Scale AI, Labelbox, and Remotasks to assess text responses and perform data labeling tasks. The focus was on enhancing model alignment and refining large language model interactions for enterprise applications. • Conducted comprehensive evaluation and annotation of AI text outputs. • Utilized multiple data labeling platforms including Scale AI, Labelbox, and Remotasks. • Maintained high standards for data accuracy in model training workflows. • Collaborated with the team to optimize prompt engineering and refine AI responses.

2023 - Present

Data Labeler and Reviewer – Labelbox

TextFine Tuning

Acted as a second‑pass reviewer in Labelbox, checking other annotators’ work, resolving disagreements, and making sure labels followed detailed project rubrics. Consistently met quality and productivity targets while documenting edge cases and giving feedback that improved guidelines and model behavio

Acted as a second‑pass reviewer in Labelbox, checking other annotators’ work, resolving disagreements, and making sure labels followed detailed project rubrics. Consistently met quality and productivity targets while documenting edge cases and giving feedback that improved guidelines and model behavio

2026 - 2026

AI Voice Assistant Data Labeler & Evaluator - LabelBox

AudioRLHF

Labeled and evaluated AI voice assistant conversations for an English‑language virtual assistant. Reviewed and corrected transcripts, tagged user intent and key entities, and rated assistant responses for correctness, helpfulness, tone, and safety. Followed detailed written guidelines, flagged policy‑violating outputs, and left clear comments on edge cases to improve both model behavior and annotation instructions.

Labeled and evaluated AI voice assistant conversations for an English‑language virtual assistant. Reviewed and corrected transcripts, tagged user intent and key entities, and rated assistant responses for correctness, helpfulness, tone, and safety. Followed detailed written guidelines, flagged policy‑violating outputs, and left clear comments on edge cases to improve both model behavior and annotation instructions.

2025 - 2026

Education

Work History