I'm a member of technical staff at Lovable, a Stockholm-based AI startup that enables anyone to build full-stack web applications using natural language. I'm also an MSc student at ETH Zurich with an AI major (machine intelligence) and data management systems minor, having recently graduated BSc in CS and math with Summa Cum Laude from École Polytechnique.
At Lovable, I'm working on democratizing software development through AI-powered tools that allow users to create production-ready websites and apps through conversational interfaces. I'm also particularly passionate about AI research in LLMs, generative AI, and large-scale data systems. My recent research focuses on LLM backdooring/model poisoning and AI for mathematics. As a math olympiad medalist, I'm excited to contribute to advancements in formal theorem proving and mathematical AI. Previously, I explored efficient Transformer design during my BSc thesis and generative seismic data modeling during an exchange at ETH.
Beyond work, I'm passionate about specialty coffee and spend time perfecting extraction techniques. I enjoy running to explore new places, traveling to discover the best food spots, reading about health and productivity (plus literature and romance novels), and watching football!
- Better LLM Backdoor - Researching stealthier triggering mechanisms and outputs in LLM backdoor attacks; conducted at the ETH SRI lab
- Project Numina - Contributor to autoformalization and formal theorem proving with LLMs
- Generative Seismic Modeling - GAN-based generation of high-quality seismic waveforms; conducted with the Swiss Data Science Center
- Prototype-based Sampling for Class-Incremental Learning - Developed a prototype-based sampling method for CIL based on the relationship between catastrophic forgetting and feature space representations
- Robustness Verification - Implemented a convex-relaxation-based verifier for neural network robustness
- Siamese Networks for Few-shot Learning - Explored few-shot learning with Omniglot dataset and compared Siamese Network architecture performance with statistical and clustering methods
- Face Identification System - Led a 13-person team building a C++ & OpenCV-based authentication system using statistical methods
- Member of Technical Staff @ Lovable - Working on AI-powered full-stack development platform that democratizes software creation
- BSc Thesis @ Huawei Paris Research Center - Efficient Transformer design research
- Research Intern @ Huawei Paris Research Center - Memory allocation optimization for parallel deep neural network training
- Quantitative Analyst Intern @ Metrix
- Math Olympiad Tutor
- ETH Excellence Scholarship & Opportunity Programme (ESOP)
- Silver Medal - International Mathematical Olympiad (IMO)
- Gold Medal - Brazilian Olympiad in Informatics
- Excellence Scholarship by the École Polytechnique Foundation