Yoni Levy

Key Skills

Software

Toloka

Appen

Data Annotation Tech

Labelbox

Micro1

Mindrift

OneForma

Remotasks

Scale AI

SuperAnnotate

Surge AI

Telus

Top Subject Matter

LLM Evaluation and AI Content

AI Model Evaluation

Adversarial AI Evaluation

Top Data Types

Text

Audio

Document

Top Task Types

Red Teaming

Segmentation

Classification

Object Detection

Text Summarization

RLHF

Fine-tuning

Transcription

Evaluation/Rating

Data Collection

Prompt + Response Writing (SFT)

Function Calling

Question Answering

Computer Programming/Coding

Freelancer Overview

AI Trainer & Evaluation Specialist — Invisible Technologies. Brings 4+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Internal, Proprietary Tooling, and Toloka. Education includes Bachelor of Arts, Reichman University (2028). AI-training focus includes data types such as Text and labeling workflows including Evaluation, Rating, and Red Teaming.

Labeling Experience

AI Red-Teamer — Mercor

MercorTextRed Teaming

As an AI Red-Teamer, I conducted adversarial testing on AI systems to reveal vulnerabilities and reasoning failures. I designed and executed prompt-based tests for cultural and safety mismatches. This role required an in-depth understanding of evaluation frameworks and model stress testing. • Probed AI for edge cases and inconsistencies • Tested safety resilience in content generation • Designed frameworks for adversarial evaluation • Identified failure modes and cultural misalignments

As an AI Red-Teamer, I conducted adversarial testing on AI systems to reveal vulnerabilities and reasoning failures. I designed and executed prompt-based tests for cultural and safety mismatches. This role required an in-depth understanding of evaluation frameworks and model stress testing. • Probed AI for edge cases and inconsistencies • Tested safety resilience in content generation • Designed frameworks for adversarial evaluation • Identified failure modes and cultural misalignments

2025 - Present

Agent Evaluation Analyst — Toloka

TolokaText

As an Agent Evaluation Analyst, I evaluated generative AI agents on reasoning, safety, and alignment using human-in-the-loop processes. I provided structured annotations and detailed feedback for direct model improvement. Consistent guideline application and calibration were integral to my responsibilities. • Evaluated generative AI agents for multiple quality dimensions • Delivered actionable feedback and annotations • Participated in calibration and QA processes • Benchmarked human evaluation against automated metrics

As an Agent Evaluation Analyst, I evaluated generative AI agents on reasoning, safety, and alignment using human-in-the-loop processes. I provided structured annotations and detailed feedback for direct model improvement. Consistent guideline application and calibration were integral to my responsibilities. • Evaluated generative AI agents for multiple quality dimensions • Delivered actionable feedback and annotations • Participated in calibration and QA processes • Benchmarked human evaluation against automated metrics

2025 - Present

AI Trainer & Evaluation Specialist — Invisible Technologies

Text

As an AI Trainer & Evaluation Specialist, I trained and evaluated large language models with a focus on reasoning, accuracy, and safety. I performed judgment-based assessment of AI-generated content, flagging semantic drift and complex annotation challenges. My work included reviewing text for consistency and guiding process improvements. • Trained and evaluated LLMs for accuracy and safety • Judged and corrected multilingual text outputs • Identified issues and improved annotation workflows • Reviewed cultural appropriateness and guideline adherence

As an AI Trainer & Evaluation Specialist, I trained and evaluated large language models with a focus on reasoning, accuracy, and safety. I performed judgment-based assessment of AI-generated content, flagging semantic drift and complex annotation challenges. My work included reviewing text for consistency and guiding process improvements. • Trained and evaluated LLMs for accuracy and safety • Judged and corrected multilingual text outputs • Identified issues and improved annotation workflows • Reviewed cultural appropriateness and guideline adherence

2023 - 2025

Education

Work History