Amit Urav

Key Skills

Software

OneForma

Top Subject Matter

No subject matter listed

Top Data Types

Text

Top Label Types

Polygon

Segmentation

Text Generation

Audio Recording

Transcription

Freelancer Overview

I have over three years of experience working with digital processes, content quality analysis, and structured learning workflows, with a strong focus on data accuracy, feedback loops, and continuous improvement. My background includes designing and evaluating AI-assisted learning modules, auditing AI-generated content for correctness and bias, and using advanced tools such as prompt-based AI systems, MS Excel, Google Workspace, and learning management systems. I am skilled at reviewing outputs for clarity and accuracy, building evaluation rubrics, and documenting quality improvements, all while working independently in remote, metric-driven environments. My hands-on experience with content annotation, quality audits, and adaptive learning systems has given me a keen understanding of the importance of high-quality training data in AI and digital education projects.

Labeling Experience

LLM Training & Evaluation – Text Data Labeling Project

OneformaTextPolygonSegmentation

Worked on a project-based AI training initiative focused on improving Large Language Model (LLM) performance. Tasks included writing high-quality prompt–response examples for Supervised Fine-Tuning (SFT), evaluating and ranking multiple AI-generated responses, providing clear justifications, and flagging factual errors, hallucinations, bias, and safety issues. Also performed RLHF-style preference rating and red teaming to stress-test edge cases and unsafe scenarios. Followed strict project rubrics, quality benchmarks, and confidentiality guidelines to ensure consistent, reliable outputs at scale.

Worked on a project-based AI training initiative focused on improving Large Language Model (LLM) performance. Tasks included writing high-quality prompt–response examples for Supervised Fine-Tuning (SFT), evaluating and ranking multiple AI-generated responses, providing clear justifications, and flagging factual errors, hallucinations, bias, and safety issues. Also performed RLHF-style preference rating and red teaming to stress-test edge cases and unsafe scenarios. Followed strict project rubrics, quality benchmarks, and confidentiality guidelines to ensure consistent, reliable outputs at scale.

2020 - 2025

Education

Work History