Pedro Henrique Bandeira Diniz

I completed two postdoctoral fellowships at the University of Oxford (Top 1 in the world for the past 10 years), in departments of the Medical Sciences Division (Top 1 in the world for the past 15 years), where I conducted research on deep/machine learning for medical imaging (2.5 years). I was supervised by leading global professors at Oxford, including the world’s Top 2 researcher in Medical Image Analysis and FSL lead (the world’s most widely used neuroscience library). I hold a Ph.D. in Computer Science, a Master’s in Electrical Engineering, and a Bachelor’s in Computer Science.

Alongside my studies, I worked in the academy and industry on Machine Learning and Deep Learning applied to several types of data (10+ years of experience). I developed what is likely the largest Medical Imaging software in Brazil and have a paper recognized as one of the best in the world in my field (competing against 1,400+ papers from 100+ journals).

Machine Learning Engineer at Eldorado Institute (2025-01 – Present)

• Developing real-time OCR solutions for Apple, using Computer Vision and Computational Geometry (Python, Tesseract, Apple Vision, and LLMs). The system currently supports more than 40 languages and is deployed in production.

Machine Learning Engineer at CERTI Foundation (2023-01 – 2025-12)

• Developed detection of ruptures in subsea petroleum ducts using Logistic Regression and LSTMs (Python, PyTorch, Scikit-learn). My method surpassed the human analyst in its first field test and is currently in deployment on the platforms of Petrobras, the largest Latin American company
• Developed full pipeline of the project, including Data Engineering and SQL Databases (Python, PostgreSQL), Cloud Development (AWS, S3, SageMaker), User Interface (Python, QT), and Software Engineering best practices (SCRUM, version control, object-oriented design, etc.)
• Developed PDF document analysis to identify and organize Petrobras equipment descriptions using LLMs (Python, OpenAI API)
• 2 patents under development

Postdoctoral Researcher at Nuffield Department of Clinical Neurosciences (NDCN), University of Oxford (2021-01 – 2022-12)

• Developed Deep/Machine Learning segmentation for ischemic and hemorrhagic stroke lesions and several head tissues in CT (Python, MATLAB, PyTorch, MONAI, etc.)
• Quality-controlled and improved a CT dataset containing 500+ subjects (MATLAB, Python) • Developed a systematic literature mapping on brain lesion segmentation by analyzing more than 100 research articles and top methods from challenges
• Developed Graphical User Interface (GUI) software for dataset visual assessment, curation, and testing of my segmentation methods (C++)
• Co-applicant for a successful grant to translate my methods to low-quality African datasets • I was supervised by the Principal Developer of FSL, the most used neuroscience library • Published 2 journal articles (unrelated to the research)

Tester (Volunteer) at DeepLearning.AI (2022-01 – 2022-12)

• Tested a Machine Learning course lectured by Andrew Ng before its public release • Provided suggestions and corrections to improve his video lectures and slides

Postdoctoral Researcher at Nuffield Department of Women's & Reproductive Health (NDWRH), University of Oxford (2020-01 – 2021-12)

• Rewrote existing Image Analysis and Deep Learning code to take advantage of updated Python libraries (Tensorflow, ITK)
• Developed a data augmentation to improve segmentation in low-quality ultrasound
• Developed a GUI software for visualization, annotation, and segmentation (C++, ITK, VTK)
• Integrated existing image analysis and deep learning tools into the GUI software
• Quality-controlled a dataset of 300 ultrasound volumes • Published 2 journal articles (1 unrelated to the research)

Professor at Instituto Federal de Educação, Ciência e Tecnologia do Maranhão (IFMA) (2019-01 – 2019-12)

• Lectured on Advanced Databases and Basic Informatics courses
• I was selected for this position in a national public contest

Research Fellow at Tecgraf Institute, Pontifical Catholic University of Rio de Janeiro (PUC-Rio) (2014-01 – 2018-12)

• Developed Deep Learning segmentation of white matter hyperintensities (WMH) in MRI (Python, Tensorflow, C++, ITK)
• Research proposed by the vice-president of DASA, the largest medical diagnostics company in Latin America
• Developed GUI software to annotate WMHs in 500+ volumes (C++, ITK, VTK)
• Pioneered the transition of Tecgraf to Deep Learning research through presentations and mentoring students
• Mentored 4 doctoral students and developed key components of their research projects, resulting in journal articles, sponsored projects, one award, and one state-of-the-art
• Published 7 peer-reviewed articles

Research Fellow at LabPAI, Applied Computing Group (NCA), Federal University of Maranhão (UFMA) (2009-01 – 2014-12)

• Main developer of SAIM, the most complex and feature-rich medical imaging software developed in Brazil
• SAIM focused on lung CT and mammography and included a large variety of visualization, segmentation, and measurement tools, including automatic lesion segmentation and cancer classification (C++, ITK, VTK, LibSVM, etc.)
• Led 3 students in a research project to detect nodules in breast thermal images (using FLIR thermal cameras) using neural networks. We also developed GUI software for annotation (C++, OpenCV, wxWidgets)
• Acted as a senior image analysis developer at LabPAI and developed many other medical research projects, and mentored and trained several co-workers
• Published 4 peer-reviewed research articles and presented them at conferences

Doctorate in Computer Science – Pontifical Catholic University of Rio de Janeiro (PUC-Rio) (2014-01 – 2018-12)

Masters in Electrical Engineering – Federal University of Maranhão (UFMA) (2012-01 – 2014-12)

Bachelor in Computer Science – Federal University of Maranhão (UFMA) (2008-01 – 2012-12)