I studied electronic engineering, computer science had just been introduced into the curriculum, and I immediately became passionate about it. My background is diverse, I also have a degree in agricultural sciences, a Cisco CCNA certification, a Microsoft certification, and I studied music and composition, participating in national and international festivals. In short, I didn't miss out on anything. Perhaps it was this varied education that allowed me to see things with different eyes.
I specialized in VoIP, becoming an expert in Asterisk and Java. I founded BeeVoip, a company with which I developed a telephone framework for creating AI agents. I studied blockchain and did applied research.
My passion for AI security began a few years ago, during an interaction with Gemini that made me realize how much these models can deviate in unexpected ways. From that moment on, I began systematically testing Claude, ChatGPT, LLaMA, and others, discovering that I could consistently circumvent restrictions. I decided to delve deeper into the architecture of these models by studying ML, Python, fine-tuning, RLHF, etc.
Today, I have developed several frameworks for LLM red teaming and mapped many of what I call “semantic attractors.” Every day, I continue to find new ways to predict and guide the behavior of these models.
Comments