Summary
Esdras Costa is a Staff Research Data Scientist with 12+ years building production-grade computer vision and deep learning systems, from foundational research to cloud and on-device deployment. With an M.Sc. focused on deep learning, meta-learning and unsupervised image anomaly detection, he blends rigorous experimental methodology with pragmatic software engineering in Agile, cross-functional teams. At CESAR he architects multi-GPU distributed training frameworks and led giga-pixel microscopy segmentation and oil & gas optimization R&D, turning complex research into deployable MVPs. He mentors researchers, communicates technical outcomes to executives and publishes pre-prints, bridging academia and industry impact. Previously he shipped large-scale web and ML pipelines at Pitang and taught hands-on deep learning courses, demonstrating a strong developer-to-instructor trajectory. Notably, he pairs advanced neural architecture work (ViTs, diffusion models, GANs) with practical optimization techniques (Bayesian optimization, PSO, GA) to solve real-world non-convex problems.
12 years of coding experience
4 years of employment as a software developer
Master of Science - MS, Computer Science - Deep Learning, Meta-Learning and Unsupervised Image Anomaly Detection, GPA: 3.7, Master of Science - MS, Computer Science - Deep Learning, Meta-Learning and Unsupervised Image Anomaly Detection, GPA: 3.7 at Universidade Federal de Pernambuco
Degree in Software Analysis and Development, Computer Science, First Class ( Awarded the Academic Laurel ), Degree in Software Analysis and Development, Computer Science, First Class ( Awarded the Academic Laurel ) at UNIBRATEC
Portuguese, English