Ygor is available for hire
Hire YgorYgor Rebouças Serpa
Verified Expert in Engineering
AI Scientist and Developer
Location
Fortaleza, Brazil
Toptal Member Since
May 30, 2022
Ygor specializes in R&D, blending a formal academic computer science education with vast professional experiences in AI, data science, game development, high-performance computing, and algorithms. Recently, he led the development of the first Brazilian computer-aided diagnostics tool for mammography images to be government-certified for commercial use. Previously, Ygor worked as an R&D professional and senior game developer on several projects.
Portfolio
Freelance
Python, PyTorch, Scikit-learn, Scikit-image, OpenCV, Flask, Unity, C#, Unity3D...
AUDO
Python, TensorFlow, Keras, Scikit-learn, Scikit-image, NumPy, Pandas, DICOM...
Onanim Studio
Unity, C#, Python, PlayFab, Steam, Unity2D, Unity3D, Games, 2D Games, Game AI...
Experience
Availability
Part-time
Preferred Environment
Windows, Visual Studio Code (VS Code), Visual Studio, Git, Unity
The most amazing...
...feat I've achieved was developing the first Brazilian computer-aided diagnostics tool for mammography images to be government-certified for commercial use.
Work Experience
Lead Data Scientist and Unity Specialist
2021 - PRESENT
Freelance
- Led the development of a computer vision system for tracking multiple persons within a gym and fitness setting; it also identifies and counts repetitions of performed exercises.
- Ported and optimized the CenterTrack architecture on Windows devices with low-end graphics processing units (GPUs).
- Developed an integration between the computer vision module and the Unity game engine for creating visually-rich representations of real-time inference.
Technologies: Python, PyTorch, Scikit-learn, Scikit-image, OpenCV, Flask, Unity, C#, Unity3D, APIs, Object Detection, 3D Pose Estimation, IP Cameras, Optimization
Lead Data Scientist
2019 - 2022
AUDO
- Developed the TensorFlow-based mammography segmentation model and related submodules for micro-calcification, malignant mass, and pectoral muscle identification.
- Optimized the developed solution to efficiently handle 16-megapixel images during training and inference and ported several algorithms to the GPU.
- Conducted the model evaluation and analysis process following the government certification process and passing all approval steps.
Technologies: Python, TensorFlow, Keras, Scikit-learn, Scikit-image, NumPy, Pandas, DICOM, Python 3
Senior Game Developer
2017 - 2022
Onanim Studio
- Developed the entire game logic for the Trajes Fatais: Suits of Fate game on Unity, including all game mechanics, asset loading, artificial intelligence, and PlayFab integration.
- Implemented a generative AI model to help artists design characters faster. The GaN-based system increased productivity by 20% and received the Best Paper Award from SBGames 2019, the Brazilian Symposium on Computer Games and Digital Entertainment.
- Built an entirely data-driven and hot reloadable system to allow game designers to edit mechanics and weights while the game is running.
Technologies: Unity, C#, Python, PlayFab, Steam, Unity2D, Unity3D, Games, 2D Games, Game AI, Game Development, Game Tools Development
Lead Data Scientist
2019 - 2021
University of Fortaleza
- Led the R&D of a human activity recognition pipeline using low-cost cameras, object detection, tracking, and state-of-the-art pose estimation technology.
- Spearheaded data acquisition, cleaning, and labeling efforts in developing a custom dataset of on-the-wild human activity videos.
- Cowrote an article, Evaluating pose estimation as a solution to the fall detection problem, published at the IEEE 8th International Conference on Serious Games and Applications for Health, SeGAH 2020.
- Managed a team of four developers, handling most of the code review and technical quality control.
Technologies: Python, TensorFlow, Keras, XGBoost, Scikit-learn, Scikit-image, OpenCV, Machine Learning, Deep Learning, JavaScript, React, Flask, Python 3
Software Engineer
2018 - 2019
TotalCross
- Took part in the platform's efforts to achieve a modern look and feel on all its supported platforms—Windows, Linux, Mac, Android, IOS, and Windows CE.
- Ported the platform's low-level rendering code from OpenGL to Skia and performed several significant optimization efforts. We observed up to 297% speed-ups across several applications.
- Performed extensive refactoring and review work on the platform's custom Java virtual machine written in pure C.
- Solved more than a hundred issues for several platform clients.
Technologies: C, C++, Java, Android, Windows, Skia, OpenGL, OpenGL ES
Researcher
2014 - 2019
University of Fortaleza
- Developed and open-sourced the Broadmark framework, a tool designed to help researchers develop, analyze, and compare novel broad-phase collision detection algorithms.
- Gathered the most exhaustive and comprehensive set of broad-phase collision detection algorithms to date and published the work in the prestigious Computer Graphics Forum.
- Created the current state-of-the-art algorithm for single-core broad-phase collision detection based on a hybrid of KD-Trees, sweep-and-prune, and temporal reasoning optimizations.
Technologies: C, C++, NVIDIA CUDA, Python, Computational Geometry Algorithms Library (CGAL), Bullet, Unity, OpenGL, Intel TBB, OpenMP, Physics, Ray Tracing
Game Developer
2014 - 2017
Valente Studio
- Codeveloped several small PC and mobile titles for several local clients.
- Coded several gameplay mechanics, UIs, and NPC behaviors.
- Performed playtests, bug fixing, and optimization activities, both on PC and on deployed mobile builds.
Technologies: Unity, Unity2D, Unity3D, Games, 2D Games, 3D Games, 3D Graphics, 2D Graphics, Graphics, C#, Game Development, Game Tools Development
Experience
Broadmark
https://github.com/ppgia-unifor/BroadmarkThe Broadmark framework was designed to help researchers develop, analyze, and compare novel broad-phase collision detection algorithms. I created the current state-of-the-art algorithm for single-core broad-phase collision detection using this framework. This single-threaded CPU algorithm outperformed all CPU algorithms and performed comparably to several GPU solutions.
Game Assets using GANs
https://github.com/nuzrub/PaintingGameAssetsWithDeepLearningThis project was a study that has shown current deep learning-based generative models are sufficiently capable of helping human artists generate art assets for games faster. In the study, we managed to speed up the drawing of game sprites up to 20% using a U-Net-based model to autocomplete drawings. This work was first published at the SBGames 2019, the Brazilian Symposium on Digital Games, where it received the Best Paper Award. Later on, we published an extended version in the Entertainment Computing Journal.
Skills
Languages
Python, C#, Python 3, Java, C, C++, SQL, HTML, CSS, JavaScript, Prolog
Frameworks
Unity, Unity2D, Unity3D, Flask, Bullet
Libraries/APIs
NumPy, TensorFlow, Keras, Scikit-learn, Pandas, Matplotlib, OpenCV, PyTorch, XGBoost, React, Skia, OpenGL, OpenGL ES, Intel TBB, OpenMP
Tools
Git, Scikit-image, Visual Studio, MATLAB, PlayFab
Other
Machine Learning, Deep Learning, DICOM, Computational Geometry Algorithms Library (CGAL), Physics, Ray Tracing, Games, 2D Games, Game AI, Game Development, Game Tools Development, APIs, Object Detection, 3D Pose Estimation, IP Cameras, Optimization, 3D Games, 3D Graphics, 2D Graphics, Graphics
Paradigms
Agile
Platforms
Windows, Visual Studio Code (VS Code), NVIDIA CUDA, Steam, Android
Education
2017 - 2019
Master's Degree in Computer Science
University of Fortaleza (UNIFOR) - Fortaleza, Brazil
2012 - 2017
Bachelor's Degree in Computer Science
University of Fortaleza (UNIFOR) - Fortaleza, Brazil
2015 - 2016
Exchange Program in Artificial Intelligence
KU Leuven - Leuven, Belgium
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring