Qk X.

Key Skills

Software

Other

Internal/Proprietary Tooling

Top Subject Matter

Cybersecurity Domain Expertise

LLM Security

AI Training

Top Data Types

Text

Computer Code Programming

Top Task Types

Prompt + Response Writing (SFT)

Text Generation

Red Teaming

Fine-tuning

Function Calling

Computer Programming/Coding

Question Answering

Freelancer Overview

AI Data Trainer specializing in LLM Alignment, Security Evaluation, and Red Teaming. Core strengths include Prompt Engineering, Vulnerability Analysis, and Adversarial Testing. Education includes a Bachelor of Science in Cybersecurity/Computer Science, Top-Tier University (Expected 2026). AI-training focus involves data types such as Text, Computer Code, and Security Scripts. Proficient in labeling workflows including Prompt + Response Writing (SFT), RLHF Preference Ranking, and Chain-of-Thought (CoT) generation.

Labeling Experience

AI Data Trainer – AI-Assisted Security Research & Vulnerability Analysis

OtherPrompt + Response Writing (SFT)

Worked on AI-assisted security research in code auditing and vulnerability analysis using generative AI models. Designed prompt-based workflows to detect and verify common web vulnerabilities, assisting AI in generating remediation suggestions. Developed and refined scripts for vulnerability assessment through AI-guided automation. • Led prompt engineering for AI in code audit tasks and vulnerability mining. • Implemented closed-loop verification processes for vulnerability remediation. • Developed automated scripts for common cybersecurity risks including XSS and SQL injection. • Integrated LLMs into security research to streamline traditional testing workflows.

2025 - Present

AI Data Trainer – LLM Alignment & Security Evaluation

OtherTextPrompt + Response Writing (SFT)

Served as an AI Data Trainer focusing on LLM alignment and security evaluation, producing high-quality prompts and responses. Developed complex instructions for cybersecurity and code vulnerability analysis to improve model behavior in security contexts. Participated in data cleaning and structuring to enhance the quality of the AI training corpus. • Crafted detailed and challenging prompts for Large Language Models within cybersecurity scenarios. • Conducted Jailbreak tests and adversarial attacks as part of red teaming for LLM security. • Utilized automation tools and proprietary/other mainstream AI software for data preparation. • Directly contributed to identifying and patching security vulnerabilities in AI systems.

2023 - Present

LLM Red Teaming

Computer Code ProgrammingRed Teaming

Constructed an adversarial test set of 300+ prompts utilizing Prompt Injection and Role-play Bypass to evaluate the refusal rates of mainstream LLMs (e.g., GPT, Claude) against malicious requests like phishing or malicious script generation. Additionally, conducted safety alignment evaluations based on the 3H principles (Helpful, Honest, Harmless), performing manual scoring and logical annotation on edge-case responses to analyze the trigger conditions and bypass patterns of the models' safety guardrails.

2025 - 2026

Education

Work History