Ronit Kumar

Key Skills

Software

Labelbox

Mercor

Scale AI

Top Subject Matter

AI Language Model Evaluation

LLM Adversarial Testing and Prompt Engineering

Top Data Types

Text

Document

Image

Top Task Types

Red Teaming

Evaluation/Rating

Fine-tuning

Freelancer Overview

AI Prompt Evaluator. professional experience across legal operations, contract review, compliance, and structured analysis. Core strengths include Mercor and Other. Education includes Bachelor of Arts, Mahatma Gandhi Kashi Vidyapith (2025) and Higher Secondary Certificate, Bhagat Public School (2022). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Red Teaming.

Labeling Experience

Independent Red Teaming Researcher

OtherTextRed Teaming

Independently designed red teaming experiments exploring adversarial and edge-case prompt patterns for GPT-4, Claude 3, and Llama models. Analyzed response behavior to adversarial scenarios, instruction injection, and multi-turn escalation. Documented and catalogued findings to understand model vulnerabilities and performance variations. • Created categorized prompt catalog documenting observed outputs and behavior differences. • Focused on role-play scenarios and instruction/context manipulation challenges. • Maintained detailed reproduction notes for all adversarial tests conducted. • Ongoing personal project tracked with regular updates and practical experiments.

Independently designed red teaming experiments exploring adversarial and edge-case prompt patterns for GPT-4, Claude 3, and Llama models. Analyzed response behavior to adversarial scenarios, instruction injection, and multi-turn escalation. Documented and catalogued findings to understand model vulnerabilities and performance variations. • Created categorized prompt catalog documenting observed outputs and behavior differences. • Focused on role-play scenarios and instruction/context manipulation challenges. • Maintained detailed reproduction notes for all adversarial tests conducted. • Ongoing personal project tracked with regular updates and practical experiments.

2025 - Present

AI Prompt Evaluator

MercorText

Evaluated and wrote prompts for GPT-4 and Claude models to assess response accuracy, safety, and tone. Documented behavior issues and flagged safety concerns in AI chatbot outputs. Provided bilingual prompt evaluation in English and Hindi, ensuring output quality across languages. • Conducted structured feedback reporting for model evaluation teams. • Identified and communicated edge cases and refusal inconsistencies in AI responses. • Used Mercor platform for performing evaluation work independently. • Summarized and submitted cycle-based written feedback on model performance.

Evaluated and wrote prompts for GPT-4 and Claude models to assess response accuracy, safety, and tone. Documented behavior issues and flagged safety concerns in AI chatbot outputs. Provided bilingual prompt evaluation in English and Hindi, ensuring output quality across languages. • Conducted structured feedback reporting for model evaluation teams. • Identified and communicated edge cases and refusal inconsistencies in AI responses. • Used Mercor platform for performing evaluation work independently. • Summarized and submitted cycle-based written feedback on model performance.

2026 - 2026

Education

Work History