I build instruments — and lately, a microscope you can afford.
Most of my work lives in the gap between a physics idea and a box that actually measures something — nanoparticles levitated in an ion trap, a spectrometer with no grating in it, a laser lock running on an FPGA. I usually end up owning the whole stack: the optics, the PCB, the firmware, and the reconstruction code on top.
I did my master's in quantum engineering at ETH Zürich, and engineering physics before that in Hong Kong. I like hardware that ships, and I'm suspicious of instruments that only work in the lab they were born in.
Now — building Eureka, a microscope with real optical quality at a price people will actually pay. Zürich. Always up for a conversation about optics, weird sensors, or starting things.
A microscope with the optical quality people expect from lab equipment, at a price they will actually pay, and small enough to carry in a bag. I design the device in Fusion 360, iterate on it with a 3D printer, and use it to capture plankton. The other half of the job is working with my co-founder to define the right product.
Can a kHz Si₃N₄ membrane be coupled to a superconducting qubit hard enough to reach deep strong coupling? Theory, an FEM model with Bayesian-optimised tether geometry, and interferometry on real metallised membranes. For a transmon, no — pull-in arrives first. For a Cooper-pair box, yes.
Built the optical bench and data pipeline, then taught a convolutional autoencoder to invert it — as accurate as linear reconstruction.
Drop a trapped nanoparticle, time the fall, get g. Designed the trap through ten revisions and debugged the high-voltage switching board.
A laser lock that beats FFT: 100 kHz resolution in a 10 µs window. Verilog through to Python — now running in several labs.
Thousands of lattice-model runs in C++ to pull a power law out of the data, and a two-level-system account that explains it.
A PID thermal package prototype, from design to bench test, plus the Ansys Icepak simulations behind a new gas-sensor product.
No framework, no build step, no dependencies. The project thumbnails are SVGs of the actual physics.
Source →An NFT collection of AI-assisted portraits, one per friend. Mostly an excuse to take a Stable Diffusion pipeline apart and read the sampler internals.
On Rarible →I wrote and produced an album about the life that starts after you leave home.
AI-assisted art, usually made as a gift rather than a portfolio piece.
3D printing, small mechanisms, and whatever needs to exist by Friday.
Easiest at hey@yuliu.dev. I answer most things, and I'm especially happy to talk about optics on a budget, instruments that escape the lab, or building a hardware team from nothing.