Looking for a job

Magdalena O’Grady

@magdalenaogrady

Multilingual AI evaluator and red teamer testing LLM safety, reasoning, and robustness in English and Polish.

In my freelance AI evaluation and adversarial testing work, I evaluate multilingual AI responses and test LLMs with jailbreaks, prompt injections, and edge cases. I report vulnerabilities, unsafe outputs, and reasoning failures through structured feedback and risk analysis.

As an AI Quality & Evaluation Analyst, I supported Polish and English model development and evaluated conversational and multimodal systems, including Grok, Gemini, and Midjourney. I also built English-language evaluation frameworks for adversarial testing and structured AI assessment.

Experience

Work history, roles, and key accomplishments

Current

AI Evaluation and Adversarial Testing

Current

Freelance

Jan 2022 - Present (4 years 8 months)

Evaluated multilingual AI responses across diverse conversational workflows and edge-case scenarios. Designed adversarial prompts to test robustness, reasoning, and instruction-following behavior.

LLM Evaluation Adversarial Testing Multilingual AI Evaluation Prompt Engineering Jailbreak Testing Hallucination Detection Behavioral Analysis Risk Reporting QA Processes

Current

AI Quality & Evaluation Analyst

Current

Contract

Jan 2022 - Present (4 years 8 months)

Evaluated and classified AI-generated content for LLM training data and quality workflows. Supported multilingual model development in Polish and English with focus on semantic consistency and reasoning quality.

AI Quality Analysis Multilingual AI Evaluation LLM Evaluation Adversarial Testing Performance Testing Behavioral Analysis QA Processes English Polish

AI Evaluator

Contract

Nov 2025 - Jan 2026 (2 months)

Reviewed and scored LLM outputs for quality, safety, factual correctness, and reasoning quality. Conducted adversarial testing including prompt injections and jailbreak chains.

LLM Evaluation Safety Validation Adversarial Testing Prompt Injection Testing Jailbreak Chains Exploit Evaluation Review Risk Reporting QA