Gabriel S. Gusmão

$ whoami

Gabriel Sabença Gusmão is a scientist, chemical engineer, and an enthusiast of scientific machine learning (SciML) and its applications to the chemical and healthcare industries. Gabriel currently serves as Chief Machine Learning Officer (CMLO) and interim Chief Technology Officer (CTO) at GlucoSense Inc., where he leads the development of real-time hybrid models for glucose metabolism and diabetes management using neural-ODEs and advanced analytics deployed over scalable cloud infrastructure.

He designed and implemented GlucoScore™, a proprietary real-time glucose stability metric, and architects the company's retrieval-augmented generation (RAG) pipeline, leveraging semantic search and curated knowledge bases to deliver personalized, actionable health insights powered by LLMs.

Gabriel holds a Ph.D. in Chemical and Biomolecular Engineering from the School of Chemical and Biomolecular Engineering (ChBE) at Georgia Institute of Technology. He was a graduate research assistant in the Medford Group under Dr. Andrew J. Medford's advisement. During his Ph.D., Gabriel developed a framework for solving high-dimensional inverse problems in transient kinetics using physics-informed neural networks (PINNs), and introduced maximum-likelihood estimators and dimensionality reduction strategies for mean-field microkinetic models. His work earned him the prestigious IBM PhD Fellowship in 2021.

Gabriel's research bridges scientific computing and chemical kinetics, enabling direct uncertainty quantification via algebraic decomposition and Fisher Information analysis. At GlucoSense, he applies these techniques to build interpretable machine learning models deployed in real-time healthcare environments.

Gabriel has over 5 years of experience in the chemical industry with Braskem (Renewables Techonologies Department), where he worked on process modeling, catalyst screening, KPIs, and advanced control using plant data from ODBC/IP21 historians. He is known for his versatility, adaptability, and the ability to translate domain knowledge into scalable ML systems.

He received his undergraduate degree in Chemical Engineering from Unicamp and completed a visiting research year at the University of California, Riverside with Prof. Phillip Christopher, working on heterogeneous catalysis using experimental and computational techniques.

Gabriel is passionate about the intersection of physics-based modeling and modern machine learning. His current challenges involve deploying PINNs at scale, addressing stiffness in dynamic systems, and ensuring interpretability in hybrid modeling frameworks. Explore his selected publications or try the interactive scientific-ML playgrounds: physics-informed inverse kinetics (KINNs), Neural-ODE structure discovery (Lotka-Volterra), and a mean-variance Neural ODE for pharmacokinetics, all running live in the browser.