Kuldeep P.

Key Skills

Software

No software listed

Top Subject Matter

LLM evaluation and safety testing

Adversarial testing and AI safety for robotics

Top Data Types

Text

Top Task Types

Red Teaming

Freelancer Overview

AI Writing Evaluator & Trainer (Outlier). Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Internal and Proprietary Tooling. Education includes Master of Arts, N/A (2021). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Red Teaming.

Labeling Experience

Red Team Specialist

TextRed Teaming

No description provided.

2025 - 2025

Red Team Specialist - Innodata

TextRed Teaming

Worked as a Red Teaming Specialist on Amazon’s upcoming intelligent motion–powered robot. Designed and executed jailbreaking prompts to identify vulnerabilities and stress-test model behaviors. Documented test results to improve model safety, robustness, and reliability through hands-on adversarial evaluation. • Crafted jailbreaking prompts to probe weaknesses • Performed hands-on testing of responses and behaviors • Identified vulnerabilities to strengthen safety and reliability • Produced reports to support safer model deployment

2025 - 2025

Red Team Specialist (Innodata on Amazon project)

TextRed Teaming

Conducted adversarial red-teaming to stress-test a soon-to-launch intelligent motion–powered robot. Designed and executed jailbreaking prompts to identify vulnerabilities in model responses and behaviors. Documented findings to improve model safety, reliability, and robustness for deployment. • Designed jailbreaking prompts to uncover vulnerabilities • Performed hands-on testing of model responses and behaviors • Stress-tested for security and safety weaknesses • Documented results to support safety and reliability improvements

2025 - 2025

AI Writing Evaluator & Trainer (Outlier)

Text

Evaluated AI-generated responses for quality, accuracy, clarity, and alignment to system instructions. Applied rubric-based judgments focused on helpfulness and factual accuracy, and scored outputs to guide LLM performance improvements. Ensured rating consistency across large batches in accordance with provided guidelines and clarified edge cases via collaboration channels. • Assessed adherence to system instructions and flagged violations • Reviewed and scored outputs to refine LLM behavior • Maintained consistent evaluations across large batches • Collaborated via Discourse to align on expectations and edge cases

2024 - 2025

Education

Work History