Bachelor thesis: GPS-denied UAV navigation
Feature-aware trajectory planning that keeps a drone over landmark-rich ground, so visual localisation stays reliable without GPS.
- Python
- SuperPoint
- PX4
- QGroundControl
- WildNav
Data science student. I build machine-learning systems, autonomous robots and the infrastructure that keeps them running.
Fribourg, CH
I'm Jérémy, known as Swaiku. I studied Data Engineering at HEIA-FR and I'm now a Master's student in Data Science at HES-SO in Lausanne, with a focus on MLOps and cloud systems. I work in French, English (B2) and German (A2).
Robots keep pulling me back in. I'm an alumnus of the Eurobot robotics competition. I wrote my Bachelor thesis on UAV navigation without GPS, and I run my own services on a Raspberry Pi 5 homelab.
What I fly with, grouped the way I reach for it.
Feature-aware trajectory planning that keeps a drone over landmark-rich ground, so visual localisation stays reliable without GPS.
First season in 2025: 2nd nationally and 6th in the international qualifications. I rebuilt the control system: Nav2 navigation, py_trees match strategy and the actuators (Dynamixel servos, linear actuators).
Running the HES-SO stand for three editions: solar protection in 2024, chronic pain and electrostimulators in 2025, Chladni figures in 2026.
The site you are on: a procedural planet generated live on the GPU, with its own Atlas of collectable worlds.
A game about catching pigeons.
Raspberry Pi 5 running my own services, my ground station.
Website and media for my archery club, where I run communications.