Embedded Systems Developer in Salmsach, Switzerland

A new DNA analysis machine developed as a cooperation between Rinco Ultrasonics and Roche Switzerland. The heart of the machine is an ultrasound generator, which aside from the main functionality, also controls all the components of the machine — motor, safety switch, door sensor, tubes sensor, and a 3x5 touch display.

I was in charge of developing software for the generator, with the following components:
• Control and regulation of the high power signal towards the ultrasound actuator. The main challenge was the physical complexity of bringing and keeping the electromechanical system in resonance.
• Control of the machine and graphical interface. The main challenge was fitting in a reduced space the control of the machine as well as a graphical interface with high-quality icons and eight languages (including Japanese) with no pre-existent driver or hardware support.
• Bootloader in the embedded system; a completely new feature for this kind of microcontroller.
• Calibration and test equipment — it required a complete change of paradigm to precisely control the rate of energy transfer to the tubes.

Ultrasound generator for a wide range of frequencies (20-70kHz) and powers (100-3,000W), completely ready for Industry 4.0.

I led the generator software development, the electromechanical system analysis, the calibration and test systems, and the continuous integration platform. Particularly challenging components were:
• Bootloaders for the two cores of the microcontroller.
• Pure software driver for a touch display, with pictures, controls, and Unicode support.
• Localization infrastructure, with a semi-automatic check of the visualization in all languages.
• Integrating webserver and webpage in the generator.
• Integrating PLC to manage the safety and multiple parallel control channels.

CAD system for the creation and grinding of high-precision tools.

I was responsible for the following components:
• Parametric and graphic definition of specific surfaces (clearances).
• Tool for the automatic definition of GUI components.
• Tool for merging localization databases.

Automatic classification of cells into three classes (cell, fragments, and fused) with a few hundred pictures labeled manually.

My responsibility was to implement the machine learning algorithm for the classification. I used a semi-supervised approach to learn from the few labeled examples and the many unlabeled ones.