During my studies at ETH Zurich, I developed a probabilistic machine learning model to predict air pollution concentrations across an urban environment using Gaussian Processes.

Unlike standard AI approaches, Probabilistic AI explicitly models uncertainty, making predictions more reliable by providing not only the estimated value but also the confidence associated with it.

I implemented a kernel-based regression framework to capture spatial correlations between sensor locations and to interpolate missing data points across the city.

The objective of this project was to explore how AI models can be made inherently reliable by integrating uncertainty quantification into their core design.

Skills used:

• Python, GPy, GPyTorch, NumPy, SciPy, Matplotlib, Scikit-learn
• Git, Github

After two participations in the Monaco Energy Boat Challenge in 2021 and 2022, the Swiss Solar Boat team set itself a new challenge — positioning the project halfway between pure optimization for competition purposes and emerging as an industrial actor in sustainable boating.

Since the propulsion system plays a critical role in the boat’s energy efficiency and performance, tests and studies were carried out to develop an optimized propulsion system. To do so, Erik Maffei and I studied the feasibility of a test bench directly integrated into a Surface Effect Ship (SES), allowing free and independent access to the testing environment.

The project consisted of several steps including system design, 3D and electronic modeling, structural analysis, and electronics integration.

Skills used:

• System design from requirements specification
• 3D mechanical modeling and 2D electronic schematics creation
• FEM analysis for structural integrity verification
• Microcontroller and electronics assembly (ESC, radio link, sensors)
• Integration of a 6-axis balance for force and torque measurement

After Swiss Solar Boat’s participation in the Monaco Energy Boat Challenge (MEBC) in 2021 and 2022, the team recognized the critical need to enhance the boat’s autonomy. This insight led them to develop the Renewable Energy Foiler (REF), a three-seater hybrid solar-hydrogen foiling boat.

In this context, I was first responsible for translating the MEBC hydrogen safety rules to the REF, thereby determining the required specifications for the ventilation system. This involved modeling the diffusion of H₂ into the boat’s compartments due to a small slit in the piping, evaluating the H₂ flux from a hazardous high-pressure leak (supersonic/compressible flow), and determining the ventilation needed to keep H₂ concentrations below the lower flammability limit.

In addition, the fuel cell required water cooling. I took part in modeling different cooling designs and translating design ideas into insightful, quantitative data.

Skills used:

• Evaluation of H₂ flux from hazardous leaks
• Ventilation design to maintain H₂ concentrations below the flammability limit
• Analysis of pressure losses, flow rates, and heat transfer across different cooling designs
• MATLAB, Microsoft Excel, compressible/incompressible fluid mechanics, heat and mass transfer

One of the team projects I completed during my Bachelor's degree at EPFL involved designing and machining a fruit gripper end-effector and detector.

This included both the ideation and performance design of the gripper, as well as the development of a system to detect berries and assess their ripeness.

Skills used:

• Python programming and OpenCV for vision and ripeness detection
• Mechanical design and machining of gripper end-effector
• CAD modeling in Fusion 360
• Version control using Git and GitHub
• Engineering design process and team collaboration