2026FeaturedSkiing IMU Data Analysis
Building a custom skiing motion sensor and analysis pipeline from scratch
PythonSignal ProcessingEmbedded SystemsMachine Learning+3
terminal
Daniel Lutziger · Portfolio─────────────────────────────Type help to get started.
›
Selected Work
Featured Projects
About
ML engineer in Zurich,
working on models that meet the physical world.
I trained multimodal transformers on sleep EEG for my master's at Idiap. These days I'm building a custom IMU sensor system to learn what a good ski carve looks like — hardware, firmware, and the model on top.
I like problems where the data is messy and the feedback loop is real: sensors, signals, embedded ML, anything where the world pushes back. I write Python and PyTorch most days and care more about understanding why something works than about the framework it's written in.
Off the keyboard I ski, lift, and occasionally write QMK firmware for keyboards that nobody asked for.
Zurich
MSc Data Science, UZH
ML · Engineering · Sensors · Embedded
01
Understand before you build
Most of the time I spend on a problem is spent reading, sketching, and being confused. The code part is usually the smallest.
02
Simple beats clever
Clever code is fun to write and miserable to maintain. If I can't explain a system in a sentence, I haven't finished designing it.
03
Models lie, data doesn't
A loss curve looks great until you plot the residuals. I trust what I can measure on real inputs, not benchmarks I picked myself.
04
Ship something rough
An ugly prototype answers questions a perfect spec can't. I'd rather have a working v0.1 with known holes than a clean v1 nobody's used.
Tech Stack
Tools I work with
ML / Deep Learning
PyTorchHugging Face Transformersscikit-learnNumPyPandasPolarsJupyter
Backend & Data
PythonFastAPIPostgreSQLSPARQL / RDFDockerREST APIs
Frontend
TypeScriptReactViteTailwind CSSFramer Motion
Infra & Tooling
LinuxHetzner VPSGitGitHub ActionsAnsibleNginx
Embedded & Mobile
MicroPythonC (BNO055 / I2C / SPI)SwiftUIRP2350 / Raspberry Pi Pico