Ricardo is available for hire
Hire RicardoRicardo Cruz
Verified Expert in Engineering
Machine Learning Developer
Porto, Portugal
Toptal member since July 25, 2022
Bio
Ricardo Cruz has proposed new deep learning and computer vision methods during his research, a PhD in 2021. His latest big job was in autonomous driving, working on RGB and LiDAR data. He enjoys a good challenge and keeping current with new technologies.
Portfolio
University of Porto [in partnership with Bosch]
Computer Vision, Deep Learning, Python, PyTorch, TensorFlow, OpenCV...
INESC TEC
PyTorch, Computer Vision, Deep Learning, TensorFlow...
Flykt
R, PHP, JavaScript, Python
Experience
- Python - 8 years
- Deep Learning - 5 years
- TensorFlow - 5 years
- Computer Vision - 5 years
- Android - 4 years
- PyTorch - 3 years
- Object Detection - 3 years
- Point Clouds - 1 year
Availability
Part-time
Preferred Environment
Linux, PyTorch, TensorFlow, Python, C++, Computer Vision
The most amazing...
...experience I've had is being selected to participate in Google Summer of Code, where I contributed to the OpenSUSE and LibreOffice open-source projects.
Work Experience
Autonomous Driving Post-Doc Researcher
2021 - 2023
University of Porto [in partnership with Bosch]
- Led a collaboration between the University of Porto and Bosch Car Multimedia to improve autonomous driving perception.
- Developed frameworks for object detection using camera and LiDAR, specifically 2D discretization and raw point clouds.
- Published new methods for efficient semantic segmentation and ordinal regression.
- Supervised six master's theses, four bachelor's projects, and other team members.
- Oversaw the HPC infrastructure with the use of Slurm.
Technologies: Computer Vision, Deep Learning, Python, PyTorch, TensorFlow, OpenCV, Artificial Intelligence (AI), High-performance Computing (HPC)
Machine Learning and Computer Vision Researcher
2015 - 2021
INESC TEC
- Focused on research and rethought fundamentals about image classification and semantic segmentation for over eight publications.
- Developed a method for background invariance using adversarial training, Used backpropagation for inference to refine existing outputs, and deployed learning-to-rank methods for class imbalance.
- Contributed to workshops, Summer School on Computer Vision (VISUM), and other events.
- Awarded "outstanding recognition" twice for organizing workshops and helping with the HPC infrastructure.
Technologies: PyTorch, Computer Vision, Deep Learning, TensorFlow, Artificial Intelligence (AI)
NLP Engineer
2014 - 2015
Flykt
- Scraped Wikivoyage to relate keywords with different city destinations. This was a non-successful startup whose goal was to search for travel destinations.
- Developed the exploratory NLP algorithm using R and also scikit-learn (Python).
- Helped to port the algorithms to PHP for deployment.
Technologies: R, PHP, JavaScript, Python
Mathematical Modelling Research
2014 - 2014
University of Porto
- Reviewed and proposed new models for the immune response after HIV infection.
- Used differential equations solved analytically using Maxima and Sympy.
- Solved some differential equations by transforming them into stochastic simulations, and the Gillespie algorithm was implemented in R.
Technologies: Python, Java, R, Differential Equations, Simulations
Experience
Light-weight Versatile One-stage Object Detection Framework
https://github.com/rpmcruz/objdetectDeveloped this one-stage object detection framework because frameworks like Detectron2 are either for two-stage models or are not versatile and simple enough to adapt to new models.
The functionality includes:
• Pre-processing.
• Supports multi-scale and anchors grids.
• Flexible model heads, training, and evaluation.
• Post-processing, such as non-maximum suppression.
• Plot utilities.
• AP metrics.
The functionality includes:
• Pre-processing.
• Supports multi-scale and anchors grids.
• Flexible model heads, training, and evaluation.
• Post-processing, such as non-maximum suppression.
• Plot utilities.
• AP metrics.
Utilities for PointNet and Similar Networks
https://github.com/rpmcruz/pnetsThese are a set of utilities for PyTorch to make it easy to do research on PointNet and similar networks.
Functionality includes:
• Augmentation methods.
• Dataset loaders.
• Metrics.
• Plotting routines.
• PointNet implementation.
Functionality includes:
• Augmentation methods.
• Dataset loaders.
• Metrics.
• Plotting routines.
• PointNet implementation.
My Android Play Store
My Play Store includes:
• A couple of games I authored.
• An object detection application I supervised bachelor's students on its development. The model runs directly on the device by exporting the model through TorchMobile.
• A couple of games I authored.
• An object detection application I supervised bachelor's students on its development. The model runs directly on the device by exporting the model through TorchMobile.
Uber Pixor Implementation
https://github.com/rpmcruz/pixorThis is an unofficial PIXOR implementation. PIXOR is a neural network for object detection in LiDAR data. It works by discretizing the point cloud onto an image (2D top view) and applying a one-stage object detection based on YOLO (but simpler, without anchors). The innovation of YOLO is that it also outputs angles.
Education
2016 - 2021
PhD in Computer Science
Joint Degree: University of Minho, Aveiro, and Porto - Portugal
2013 - 2015
Master's Degree in Informatics and Applied Mathematics
Faculty of Sciences, University of Porto - Porto, Portugal
2009 - 2012
Bachelor's Degree in Computer Science
Faculty of Sciences, University of Porto - Porto, Portugal
Certifications
AUGUST 2017 - PRESENT
Deep Learning Specialization
Coursera
MARCH 2014 - PRESENT
Machine Learning Specialization
Coursera
Skills
Libraries/APIs
PyTorch, TensorFlow, OpenCV
Tools
MATLAB
Languages
Python, R, C, C++, Java, PHP, JavaScript
Platforms
Linux, Android
Frameworks
Flask
Storage
Databases
Paradigms
High-performance Computing (HPC)
Other
Computer Vision, Deep Learning, Machine Learning, Object Detection, Statistics, Optimization, Web Development, Artificial Intelligence (AI), Applied Mathematics, Point Clouds, Generative Adversarial Networks (GANs), Differential Equations, Simulations
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