My first steps into AI were as an undergraduate, when I was fortunate enough to be selected for an exchange research program with the Real Time Computer Vision Team at the Institute of Neural Computation at Ruhr-Universität in Bochum, Germany.

Later, I delved into the safety concerns surrounding genomic data, which became the focus of my master’s thesis. My goal was to better understand how FASTA files could be compromised by adversaries and how these vulnerabilities could impact digital-biological systems. Interestingly, some of my hypotheses introduced in 2015 were later confirmed in Compromising Computers with Synthesized DNA by Key et al. (2017).

I soon realized that my primary interest lies in doing interdisciplinary work. This led me to pursue graduate school to deepen my understanding of AI Safety, a topic that blends multiple areas into a challenging, emerging field.

PhD in AI Safety

I completed my PhD at the Research Institute CODE in Munich, where I had the privilege of working under the guidance of Prof. Dr. Gabi Dreo Rodosek from the Universität der Bundeswehr and Prof. Dr. Lorenzo Cavallaro from University College London, UK. My research focused on the trustworthiness of AI, specifically exploring how machine learning models respond to carefully crafted adversarial manipulations and how to improve their resilience against adaptive adversaries.

The book based on my dissertation is available on Amazon, and the accompanying software package can be installed via pip.

AI at Google X

During my PhD, I also had the incredible opportunity to join the AI Residency program at Google X in Mountain View, California where I worked on improving the performance of a challenging real-world problem. Following this, I returned to Google X to initiate a new project and eventually became one of the early founding members of SandboxAQ, a spin-off company that integrates artificial intelligence with quantum technologies.

Current work

Currently, I am a Staff Machine Learning Engineer at the AQMed team of SandboxAQ, where I work on detecting cardiovascular diseases using a cutting-edge magnetocardiography (MCG) device. Additionally, I am the author of the FAME library, which is designed to assess ML-based classifiers against adversarial attacks with malicious software. I also support early-stage startups, providing guidance on defining AI products effectively.

Origins

It’s often jokingly said, that nobody really knows where I’m from. There’s some truth to this throughout my life, I’ve had the privilege of living in five different countries: Brasil, Uruguay, Argentina, Germany, and the United States. Each of these places is unique in its own way, yet what always stands out to me is how similar people are regardless where they come from.

Fun

In my free time, I am an avid traveler (visited 50+ countries) and a passionate photographer who is always excited about the next moonshot project 💡🚀