Youssef Z.

Key Skills

Software

Don't disclose

CVAT

Other

Top Subject Matter

Cybersecurity Domain Expertise

LLM response evaluation

RLHF & AI Red Teaming

Top Data Types

Text

Document

Top Task Types

RLHF

Prompt + Response Writing (SFT)

Red Teaming

Classification

Freelancer Overview

AI training and LLM evaluation labeling (LLM response rating and preference ranking). Brings 5+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other, Don't disclose, and CVAT. Education includes Bachelor of Science, N/A. AI-training focus includes data types such as Text, Computer Code, and Programming and labeling workflows including Evaluation, Rating, and RLHF.

Labeling Experience

Senior Offensive Security Engineer / Red Team Specialist | Done Official

CVATTextRed Teaming

Conduct authorized security assessments across web applications, APIs, cloud-adjacent services, internal environments, and identity/access-control workflows. Validate vulnerabilities from discovery through impact proof, business-risk explanation, remediation guidance, and retest-ready reporting. Review technical evidence, logs, requests/responses, screenshots, and code behavior to distinguish real risk from false positives. • Build automation and reporting workflows that structure findings, severity, reproduction steps, proof, and recommended fixes. • Translate offensive-security concepts into clear instructions, acceptance criteria, checklists, and evaluation rubrics. • Communicate results for both technical and executive audiences with evidence-backed judgments. • Apply deep knowledge of web/API security, authentication/authorization testing, and risk validation.

2025 - Present

Senior Offensive Security Engineer and Red Team Specialist - Done Official

CVATRed Teaming

Led authorized offensive security engagements focused on web, API, and access-control workflows, ensuring vulnerabilities are validated with clear evidence. Owned the full assessment lifecycle from discovery and impact proof through business-risk explanation, remediation guidance, and retesting-ready reporting. Applied security automation and structured documentation to translate offensive findings into actionable technical and executive outputs using clear evaluation criteria. • Conducted security assessments across web applications, APIs, cloud-adjacent services, internal environments, and identity/access-control workflows. • Validated vulnerabilities end-to-end using evidence, logs, requests/responses, and observed code behavior to reduce false positives. • Built repeatable reporting workflows covering severity, reproduction steps, proof, and recommended fixes. • Translated offensive-security concepts into checklists, acceptance criteria, and rubric-style guidance for consistent review and validation.

2025 - Present

Senior Offensive Security Engineer / Red Team Specialist — evidence-based labeling & rubric generation for AI training

Don't discloseTextRed Teaming

Performed reviewer-style judgment on cybersecurity findings by distinguishing true risk from false positives using evidence from requests, logs, screenshots, and code behavior. Translated penetration testing observations into structured acceptance criteria, checklists, and evaluation rubrics aligned with LLM assessment needs. Produced evidence-based explanations and remediation guidance that can be repurposed into labeled training/evaluation content. • Validated vulnerabilities from discovery through impact proof and retest-ready reporting • Reviewed technical evidence and mapped risk severity using standardized assessment practices • Converted findings into structured artifacts for technical and executive audiences • Generated repeatable guidance that supports rubric-driven evaluation and dataset QA

2025 - Present

Cybersecurity annotation labeling (vulnerability class/severity/exploitability/remediation) for AI training datasets

Don't discloseTextClassification

Annotated cybersecurity data by labeling vulnerability class (OWASP/CWE-style) and extracting root cause information for classifier training. Determined exploitability using attacker path and practical impact reasoning, and provided severity rationales based on impact, likelihood, affected scope, and evidence quality. Labeled remediation quality by assessing specificity, feasibility, and secure coding accuracy to support evaluator datasets and incident-style analysis. • Vulnerability class and root cause labeling for OWASP/CWE classifier training • Exploitability and practical impact reasoning for risk ranking • Severity rationale labeling with evidence quality and affected scope • Remediation quality evaluation for answer grading and secure coding accuracy

2022 - Present

RLHF preference ranking and schema-aware labeling QA for cybersecurity datasets

OtherRLHF

Performed RLHF-style preference and grading by comparing two or more AI answers and selecting the stronger output using justified evaluator rationales. Verified that structured labeling outputs matched expected schema categories and formatting for reliable dataset ingestion. Ensured evaluation labels supported answer grading, prompt-response ranking, and high-precision QA review for cybersecurity domains. • Preference ranking to support RLHF training signals • Schema-aware output review for category consistency and edge cases • Answer grading with severity and remediation rationale scoring • Data QA consistency checks to prevent duplicate labels and severity inflation

2022 - Present

Show more

Education

Work History