Zurich, Zurich, Switzerland
Versatile embedded systems engineer with experience in multidisciplinary R&D projects across the medical device and aerospace industries. Diverse skill set spanning bare-metal embedded firmware development, CAD/CAE, machine learning and rapid prototyping. Eager to drive forward innovative R&D projects in fast-paced, results-driven environments to bring impactful technologies from concept to market.
Developing embedded firmware and mobile application for a neurostimulation wearable device for sensory feedback restoration in people with diabetic neuropathy. Led the development of the initial IP strategy in CH and the US, including patent ideation and prior art search. MYNERVA is a medical device spin-off from ETH Zürich funded by the Wyss Zurich accelerator. • Architected and built embedded C firmware on the Nordic Semiconductor nRF5340 SoC to control the Leia hardware and interface with the mobile application via Bluetooth Low Energy. Features include: - DPPI-driven stimulation pulse edge generation - Data collection and processing (stimulation waveforms, pressure sensors, battery monitoring, ...) - Bluetooth Low Energy for wireless interface with mobile application - State and configuration management - Firmware-Over-The-Air Updates - Exhaustive debug logging - Documentation following IEC 62304 standards for class B medical devices • Architected and built a Flutter mobile application for iOS and Android devices to serve as end-user controller of Leia: - Authentication - User profile configuration - Sensory feedback calibration algorithm design and implementation for the Leia device - Bluetooth Low Energy interface and Leia device connection lifecycle handling - Firestore & Firebase for cloud data storage • IP & Regulatory activities: - Conducted Freedom-to-Operate analysis in both CH and the US. - Submitted one patent application for Leia (U.S. Application No.: 19/232,065). - Registered both figurative and word trademarks with the IPI. - Assisted with the preliminary development of the US regulatory pathway (510k).
Title: Developing a Realistic Computational Model For Vagus Nerve Stimulation Starting point was an existing biophysical computational model for cervical vagus nerve stimulation (illustrated below). It could determine how much charge has to be injected by a monopolar square pulse in order to recruit large myelinated fibers. • Integrated histological data of precise fiber placements and designations as well as a computational model for unmyelinated fibers. • Built a parametric generator to easily synthesize arbitrary multipolar stimulation policies. • Developed a novel method (dynamic discretization) to more efficiently simulate unmyelinated nerve fibers, reducing the required computation time by an order of magnitude while inducing less than 0.5% error. • Developed an additional method (longitudinal truncation) to more efficiently simulate all nerve fibers, reducing the required computation time by up to an order of magnitude while inducing less than 0.5% error. • Further optimized the pipeline and implemented various technical improvements, such as leaner code and high-performance cluster support, which combined with the previously mentioned algorithms boosted the computational efficiency of the pipeline by three to four orders of magnitude in total. • Showed that considering the 3D structure of fascicles, including merging and branching, causes the estimated recruitment thresholds to vary by about 15% compared to a model with linearly extruded fascicles. • Showed that knowledge of fiber clustering, if present, can significantly improve the selectivity of tailored stimulation policies synthesized using the computational model. • Showed that variations of location and rotation of a helical cuff electrode with 8 contacts during surgical placement do not significantly effect the achievable fascicular selectivity. Here 18 different electrode configurations were evaluated. All previously mentioned findings have been verified using two distinct nerve models.
Title: Single Axon Physiology on CMOS Based Microelectrode Arrays • Designed PDMS structures to enable the electrical recording and stimulation of neuronal networks grown on CMOS microelectrode arrays at single axon resolution • Seeded, imaged and electrically recorded neural cell cultures and processed spike data
Title: Exploiting Temporal Dynamics of Spiking Autoencoders for the Processing of Biosignals • Investigated a novel autoencoder-based spiking neural network architecture for the processing of biosignals in the context of actuating a robotic hand using Python and Brian2
Title: Development and Evaluation of a Coma Free Secondary Mirror Mechanism Concept for SOFIA Grade: 1.0/1.0 - Very Good with Distinction • Conceptualized and evaluated next generation coma-free secondary mirror chopping mechanisms for the airborne SOFIA (Stratospheric Observatory For Infrared Astronomy) telescope, yielding the concept of an extremly compact spherical parallel manipulator • Derived and analyzed the full kinematic model of this design • Optimized the geometric parameters for accuracy using a genetic algorithm in Matlab • Designed and implemented flexure joints for improved precision and reduced abrasion