How I Work
I build perception-driven robotics software that has to perform outside the lab.
I am based in Zurich. I currently work at P8S, building warehouse automation with mobile robots, smart containers, and AI software. My focus is practical robotics and perception software with production-grade reliability. Before this, I worked at Magic Leap on headpose and relocalization, and at Rovenso (now Wayout Robotics) on navigation, control, and sensor fusion for autonomous ground robots.
I like taking systems from idea to release — starting with clear metrics, improving failure behavior, and keeping the codebase maintainable for the team that runs it.
How I Can Help a Team
- I build SLAM, relocalization, and sensor-fusion modules that stay reliable in real operating conditions, not just demos.
- I turn algorithms into production code with clear interfaces, measurable KPIs, and repeatable validation.
- I can work across the stack: perception modules, hardware-facing drivers, CAD mounts/fixtures, data processing pipelines, and frontend/operator tooling.
What I Build
Perception, SLAM & Sensor Fusion
Visual-inertial SLAM, relocalization, pose estimation, and multi-sensor fusion for robotics and AR systems.
AI/ML Inference
Learning-based perception modules and inference pipelines that can run within real-time product constraints.
Product Integration & Engineering Tooling
Service layers, KPI/telemetry pipelines, and operator-facing tools that help teams ship and support systems in production.
Core Stack
Also used in robotics estimation work: GTSAM, Ceres Solver, Eigen, OpenGV, PCL, RViz.
Engineering Approach
- I work from measurable behavior first: define expected latency, stability, and failure modes, then validate against data.
- I design for operability, not just demos: clear logs, reproducible runs, versioned configs, and straightforward rollback paths.
- I prefer small, testable increments that reduce integration risk and keep system behavior understandable as complexity grows.