Matheus Kunzler Maldaner

Overview

I worked with Drs. Damon Woodard and Domenic Forte conducting research on Neurosymbolic AI, combining symbolic reasoning and neural networks to create more transparent and explainable AI models. This research bridges the gap between high-level symbolic reasoning and low-level neural computation.

Key Contributions

• Explainable AI Framework: Contributed towards a novel framework for Explainable AI that fuses symbolic reasoning with neural network methodologies, enhancing the transparency and interpretability of AI systems.
• Large-Scale Model Training: Trained 260 Differentiable Logic Gate Networks on HiPerGator, UF's high-performance computing cluster, to develop and validate new approaches to transparent AI.
• Research Presentations and Publications: Presented 9 research posters at various academic venues, second-author paper accepted to Springer AI, and contributed novel findings towards ITNG '25 paper.

Research Impact

This research culminated in a Summa Cum Laude undergraduate honors thesis focused on Differentiable Logic Gate Networks and their applications. Key achievements include developing visualization software to understand network connections and learned logic patterns, and successfully extracting learned gates to deploy on FPGAs, demonstrating significantly higher inference speeds compared to traditional neural network approaches. This work advances Neurosymbolic AI by providing both theoretical frameworks and practical tools for creating more transparent, efficient, and explainable AI systems.