Jakub Sochor

Professional Experience

I was a team leader of a Fingerprints Team with 7 team members
The team worked on fingerprints verification, identification, detection, etc.
While I led the team, multiple novel solutions were implemented and deployed - for example:
- Fingerprints verification heavily utilising neural networks outperforming all the other industry solutions as verified by the first place in NIST MINEX benchmark
- Fingerprints identification based on neural networks outperforming the previous solution both in accuracy and speed (100x speed up)
I also led infrastructure, quality checks, and shared code improvements across the whole R&D deparment

I was a member of Proactive Google Assistant
I owned and led various projects that applied Machine Learning methods
Other, more junior, team members have worked on the projects that I have led; therefore, I was responsible also for the coordination of the project work and communicating the status to stakeholders
The features I have launched are used by hundreds of thousands of users daily

I was a key team member working on a project that used Machine Learning methods for Augmented Reality
The goal of the project was to estimate 6 DoF pose of objects without any real training images using only 3D models of the objects
I was responsible for the initial design, implementation, and experimentation, while the other team members started contributing to the project afterwards
The project was commercially launched as a part of the Vuforia Engine for Augmented Reality
See a demo of the project and a reference letter

I spent three months on a secondment in Vlatacom within iARTIST European project
The outcome of the secondment was a proof of concept for the application of fine-grained vehicle classification and vehicle colour recognition

I worked on research in automatic traffic surveillance, specifically vehicle re-identification, automatic camera calibration, fine-grained recognition of vehicles
The results of the research were published in top Computer Vision and Machine Learning journals and conferences (CVPR, CVIU, etc.)
The contributions to the State of the Art included improved design of Convolution Neural Networks, automatic camera calibration methods, and other

I worked on a project that was focused on testing aircraft hardware and software using Computer Vision methods