Hadrien Ovazza

CADLM is a company specializing in the optimization of structural calculations. During my permanent contract, I developed numerous applications for our clients as well as the company's entire software suite. Here are some of the tasks performed:

Ariane group diagnoses its rocket engines using accelerometers. In this context, I developed software for analyzing spectrograms from vibration recordings. Coded in C++ with the Qt framework for its graphical interface, this software stands out for its complete and intuitive spectrogram browser. I also implemented a sensitive and efficient line extraction algorithm, as well as a Python library to simplify data export for the client.

We were involved in the DGA's Man Machine Teaming project to develop intelligent compression for aerial imagery. The C++ algorithm I developed, consisting of a wavelet decomposition, arithmetic compression, and interest point detection, distinguished itself from others for its compression ratio and restitution quality. I also created an automatic testing bench in Quasar to simplify algorithm evaluation.

CADLM conducts numerous structural calculation optimization studies for its clients. I developed the Quasar2 scripting language to facilitate data exchange between simulation software and optimization algorithms. This software, inspired by several languages, is as simple as Python, as flexible as C++, facilitates matrix calculation like Matlab, and allows launching applications with shell syntax. Furthermore, my remarkable efficiency allowed the software to be released in just 3 months!

When doing optimization, we work with cost functions dependent on many variables that are impossible to visualize on a 2D or 3D graph. I internally developed the Supernova software for CADLM, which offers to display these functions in batches of slices controlled by cursors representing each variable. Developed in C++ with OpenGL for smooth rendering, this software has been used to visualize the effect of complex interpolators as well as to explain optimization mechanisms to novices.

Vedecom collects petabytes of driving data to reconstruct driving scenes for training autonomous vehicle algorithms. On their behalf, I developed the main algorithm in Matlab for fusing data from various sensors: lidar, radar, image processing camera, accelerometers... The client appreciated the precision and speed of my algorithms as well as the quality of my presentations. I also implemented a C++ and OpenGL driving scene visualization software that proved very useful throughout the project.