About
I’m a Machine Learning Engineer with a broad background in Machine Learning and Deep Learning, and specilized in Computer Vision and 3D geometry. Beyond training models, I have experience building clean, maintainable systems with solid architecture and CI/CD practices. Having worked closely with graphics engineers, cloud and infrastructure teams, and 3D artists, I have developed a strong understanding of how to communicate effectively across different domains to build real-world systems. My work has led to two SIGGRAPH publications and contributions to multiple granted US patents.
- Age: 30
- City: Porto, Portugal
- Degree: Master's
- Email: [email protected]
Technical Skills
Programming languages
- Python
- C++ (OpenCV, FBX SDK)
- C#
- Matlab
Engineering & Tooling
- Git, Docker, Linux
- Amazon Bedrock
- LaTeX
Libraries
- PyTorch
- OpenCV, Scikit-Image, Pillow, MediaPipe
- Open3D, Trimesh, PyMeshLab
- NumPy, SciPy, Pandas, Scikit-Learn
- LangChain/ LangGraph, Docling
Machine Learning
- Deep Learning for vision: CNNs, GANs, Diffusion Models, Autoencoders, ViT, CLIP
- Computer Vision tasks: segmentation, keypoint detection, regression, classification, and style transfer
- Synthetic data generation
- Geometry/ Graphics algorithms: collision detection, decimation, voxelization, texture packing
- AI Agents & RAG
3D & Graphics
- Blender - global tool usage & scripting
- Unity - scripting in C#
- MetaHuman Creator
Languages
- Portuguese (native)
- English (fluent)
Resume
Education
Integrated Master's in Bioengineering
2014 - 2019
Faculty of Engineering of the University of Porto (FEUP)
- Specialization in Biomedical Engineering.
- Master thesis: Thoracic MRI emulation through texture synthesis and ground truth manipulation. Thesis developed at the Institute for Systems and Computer Engineering, Technology and Science (INESC TEC) and published here.
For my Master's thesis, I focused on using Generative Adversarial Networks to generate synthetic thoracic MRI data from labeled images. The objective was to create novel, synthetic MRIs that could be used to train other Machine Learning algorithms in the medical field (e.g. segmentation of bones and tissues, tumor detection, etc).
Professional Experience
Senior Software Engineer
August 2019 - Present
Didimo, Porto, Portugal
Didimo is a company that offers two primary services:- creation of high-quality and realistic 3D avatars from a single facial photo;
- automated creation of non-playable characters (NPCs) for the gaming industry, with variety and at scale
I currently work on the core engine that generates the avatars, specifically on facial reconstruction and conditional face generation. Most of my work focuses on the Computer Vision challenges required for the generation of realistic and clean facial assets (geometry and textures) and on integrating those developments mindfully into the product. Working alongside 3D artists and Unity/ Unreal developers, I have also had the opportunity to tackle Computer Graphics challenges, having worked on a mix of topics from texture packing to rendering, among many others. I advocate for clean, efficient, tested, and overall good-quality code, so I also dedicate some of my time to software design challenges.
At Didimo, I had the opportunity to contribute to the development of a product from its early days. I learned that resilience, patience, teamwork, and knowledge sharing are at the core of producing good-quality software. I experienced what it is like to be at the beginning of a team and to welcome and integrate new people into it, which reinforced my appreciation for mutual support and collective growth.
Internship
September 2017 - June 2018
Institute for Systems and Computer Engineering, Technology and Science (INESC TEC), Porto, Portugal
Intern of the Breast Research Group (BRG), at the Telecommunication and Multimedia Center (CTM), at INESC TEC. Developed Computer Vision algorithms to perform multiple tasks with medical images, including segmentation and image generation, using thoracic MRI data.
Publications
Creating Infinite Characters From a Single Template: How Automation May Give Super Powers to 3D Artists
SIGGRAPH 2024
Mariana Dias, Pedro Coelho, Rui Figueiredo, Rita Carvalho, Verónica Orvalho, and Alexis Roche.
In ACM SIGGRAPH 2024 Talks (SIGGRAPH '24). Association for Computing Machinery, New York, NY, USA, Article 47, 1–2.
DOIHigh-fidelity facial reconstruction from a single photo using photo-realistic rendering
SIGGRAPH 2022
Mariana Dias, Alexis Roche, Margarida Fernandes, and Verónica Orvalho.
In ACM SIGGRAPH 2022 Talks (SIGGRAPH '22). Association for Computing Machinery, New York, NY, USA, Article 17, 1–2.
DOIAdversarial Data Augmentation on Breast MRI Segmentation
2021
Teixeira, J.F.; Dias, M.; Batista, E.; Costa, J.; Teixeira, L.F.; Oliveira, H.P.
Appl. Sci. 2021, 11, 4554.
DOIAutomatic Sternum Segmentation in Thoracic MRI
2019
M. Dias, B. Rocha, J. F. Teixeira and H. P. Oliveira
41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Berlin, Germany, 2019, pp. 1018-1021
DOI
