Radu Sibechi
Verified Expert in Engineering
Machine Learning Researcher and Developer
Radu is a passionate machine learning researcher with a strong background in programming and mathematics who has developed multiple novel projects at the intersection of computer vision and deep learning. Radu is interested in complex tasks related to various aspects of computer science, specifically in deep learning and computer vision. The new doesn't scare him but instead, he views it as a pool full of opportunities.
Portfolio
Experience
Availability
Preferred Environment
Git, PyCharm, Linux
The most amazing...
...thing I've built was a novel approach for video segmentation based on both labeled and unlabeled data, which was accepted at a top computer vision conference.
Work Experience
Senior Machine Learning Research Engineer
Stickermule
- Developed and deployed a background removal project based on F3Net.
- Built and deployed an image generation project based on Stable diffusion.
- Constructed and deployed an image impainting algorithm based on research paper ZITS.
Senior Machine Learning Research Engineer
Quin
- Led all the architectural discussions, design, and risk storming sessions.
- Designed and implemented a framework for large-scale training of machine learning algorithms.
- Prototyped deep learning models for digitalizing care pathways and designed and implemented services to productize them.
Machine Learning Research Engineer | Consultant
SevenFifty
- Implemented an automatic circular deformation applied to the flat label to match the given bottle shape.
- Infused the lightning shading from the bottle and applied it to the label for a more realistic outcome.
- Automatically detected years from labels and painted them with the background—CRAFT-PyTorch for text detection and OCR.
Machine Learning and Computer Vision Team Lead (Consultant)
Clearquote
- Oversaw the overall architecture/solution for problems faced by the company.
- Assessed the current architecture and provided feedback on areas for improvement.
- Suggested and analyzed approaches for resolving issues such as light reflectance, working with limited annotated data, increasing video performance, and so on.
Machine Learning Research Engineer
Prime Vision
- Collaborated with other computer vision experts within the team to implement state-of-the-art computer vision and machine learning solutions for tasks such as object character reading, object detection, and image/video segmentation.
- Implemented and extended a state-of-the-art approach for real-time image/document unwarping by predicting 2D deformation fields, which are synthetically generated.
- Spearheaded the rewriting of a legacy, highly-coupled core project in order to tackle technical debt.
- Worked as part of the team tasked with creating the overall architecture for detecting, tracking, and reidentifying parcels in a delivery van.
Machine Learning Research Intern
TomTom
- Researched a deep, end-to-end novel model to leverage temporal information in order to make use of easy-to-acquire unlabeled data next to labeled data for semantic segmentation.
- Implemented several recent, state-of-the-art papers in order to benchmark results.
- Wrote a scientific paper that was accepted in the CVRSUAD 2019 workshop at ICCV.
- Extended image segmentation models to efficiently work on videos with the same amount of annotations.
Medior Software Engineer
Virtuagym
- Coordinated and participated in the rewriting process of a highly coupled legacy component.
- Integrated with various payment providers to accept all major payment options.
- Developed the backbone of the API platform to be consumed by the application microservices.
Experience
Exploiting Temporality for Semi-supervised Video Segmentation
https://github.com/mhashas/Exploiting-Temporality-For-Semi-Supervised-Video-SegmentationExploring Model Architectures for Visual Question Answering
https://github.com/mhashas/Exploring-model-architectures-for-Visual-Question-AnsweringNeural Machine Translation with Attention and Positional Embeddings
https://github.com/mhashas/Neural-Machine-Translation-with-Attention-and-Positional-EmbeddingsIn this project, we attempted to solve the NMT by using a model inspired by the sequence-to-sequence approach. However, instead of encoding the sentence using a recurrent neural network, we made use of positional embeddings. We displayed our findings and drew meaningful conclusions from them.
Real-time Document Unwrapping Using MobileNet and DeepLabv3+
https://github.com/mhashas/Document-UnwarpingIn this project, we explored multiple convolutional neural network architecture types such as DeepLab, UNet, and PSPNet to predict a deformation field that can be used to unwrap the deformed input image. We made use of the novel loss function implemented in the scientific paper DocuNet™.
Education
Master's Degree in Artificial Intelligence
University of Amsterdam - Amsterdam, Netherlands
Bachelor's Degree in Computer Science
Vrije Universiteit Amsterdam - Amsterdam, Netherlands
Skills
Libraries/APIs
PyTorch, NumPy, TensorFlow, Scikit-learn, Keras, Matplotlib, Pandas, REST APIs, OpenCV
Tools
Git, PyCharm, MATLAB, LaTeX, Shell
Languages
Python, PHP, JavaScript, SQL, HTML, CSS, Python 3
Paradigms
Test-driven Development (TDD), Unit Testing, Microservices, Data Science, REST
Platforms
Linux, Windows
Other
Machine Learning, Computer Vision, Deep Learning, Data Visualization, Software Architecture, LSTM Networks, OCR, Neural Networks, Artificial Intelligence (AI), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNNs), Gated Recurrent Unit (GRU), Supervised Learning, Unsupervised Learning, Natural Language Processing (NLP), Image Recognition, Statistical Modeling, Support Vector Machines (SVM), APIs, GPT, Generative Pre-trained Transformers (GPT)
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring