About

Background and approach of Rafael Testa, an Applied AI and Machine Learning Engineer and researcher.

I am an Applied AI and Machine Learning Engineer with a PhD in Information Systems and more than ten years of experience across AI research, software development, data engineering, experimentation, and research infrastructure. I focus on problems where model performance alone is insufficient. Structure, temporal behavior, validation, and evaluation often determine whether an AI system is useful in practice.

My research background is in computer vision and generative models. I completed my BSc, MSc, and PhD at the University of São Paulo. During my doctoral research from 2019 to 2024, I studied facial-expression synthesis in video. I found that plausible individual frames could still produce visible flicker and unstable motion. I therefore developed methods based on facial geometry, temporal conditioning, pixel-transition consistency, and video-oriented evaluation. From 2021 to 2022, a Fulbright Doctoral Dissertation Research Award supported my work as a Visiting Researcher at the University of Southern California on lightweight facial-expression classification and accuracy-complexity trade-offs.

Today, I work as an AI/ML Engineer on a FUSP and Ministry of Agrarian Development project. I lead the implementation and evaluation of an internal natural-language analytics platform. My work covers stakeholder requirements, data ingestion, PostgreSQL modeling, ontology and schema grounding, Text-to-SQL, backend services, validation, security controls, deployment, and operational support. The platform treats model-generated output as untrusted until independent checks establish that it is structurally valid, grounded in the intended data, and safe to pass to the next stage.

I have also worked as a research and innovation specialist supporting shared GPU and virtual-reality infrastructure, and as an engineering professor and AI research supervisor at Centro Universitário FEI and UNIVESP. In these roles, I supported experiments, taught computing and AI subjects, and supervised projects from problem formulation through implementation, evaluation, and technical reporting.

I am interested in roles and collaborations that combine machine learning, software engineering, data systems, and rigorous evaluation. My full professional history is maintained on LinkedIn. My publication record is available through Google Scholar and ORCID.