Why PreFab exists

A photonic device design can look promising in simulation and still perform differently once it’s fabricated.

In our inverse-design work, we saw it firsthand: designs that followed the same fabrication rules could have very different measured performance. Even identical designs could behave differently because of random fabrication variations.

Learning from fabrication

Design rules gave us useful boundaries, but they couldn’t tell us the whole story. We needed to understand not just whether a design could be accepted, but how fabrication would affect its performance.

That changed the question we were asking: what can a fabrication process reliably produce, and how could we account for that while designing?

Machine learning offered a way to learn directly from fabrication data, including effects that are difficult to capture in rules or correction tables.

But studying a process and building tools around it takes time and resources. Every photonics team shouldn’t have to do that work from scratch.

Making room for more ideas

We started PreFab to make that understanding available: tools to help engineers anticipate fabrication effects and account for them before making a device.

Our goal is to give engineers more confidence in how their devices will perform after fabrication, whether they’re exploring a new idea or working with an established design. We want fabrication knowledge to inform everyday engineering decisions, not just the next experiment.