I will evaluate your llm chatbots or ai agent for bias, safety, robustness and more

Your chatbot works in the demo. But how does it behave when users push it?

I'm a data science engineer (Grenoble INP ENSIMAG, applied mathematics and AI) with hands-on experience evaluating large language models and AI agents at a major European bank. I test LLM applications the way a risk team would, before your users or clients find the problems.

What I evaluate

• Bias and unfair treatment of users
• Toxic or inappropriate outputs
• Privacy leakage of sensitive data
• Robustness to tricky or adversarial prompts
• Sycophancy: agreeing with users when it shouldn't

What makes this different

I test in English, French and Arabic, so multilingual products are evaluated in the languages your users actually speak.

What you get

A clear report with scores, concrete failure examples and prioritized fixes. With Premium, an automated Python pipeline you can rerun after every model update.

Works with GPT-based apps, open-source models, RAG chatbots and agentic systems.

Message me before ordering so we can confirm scope and access.