What is Qfactr?
Qfactr is a photonic integrated circuit design workspace and representation layer — a visual, editable, physically grounded environment for building, exploring, and communicating photonic systems.
Qfactr is an early-stage company building physics-informed generative ML for photonic integrated circuit (PIC) layout, with the goal of democratising photonic chip design. The product is a design workspace and representation layer: a place to build, route, simulate-check, and hand off photonic circuits without leaving a single editable model.
It is built on top of the open-source Lunima engine. Lunima is the open core; Qfactr is the company and product built on it. This page frames what Qfactr is, the idea behind it, and how the rest of these docs are organised.
The core idea
Photonic design should be physical, explicit, and simulation-aware. In Qfactr, components live in real micrometers (µm) rather than abstract grid tiles — the geometry is true from the first placement. Waveguides are routed as real S-bend and Manhattan paths between real component pins, not drawn as abstract nets. And because transmission and loss are derived from the actual path geometry, the model already knows enough physics to tell you what a route costs.
This is a deliberate contrast with schematic-only and spreadsheet-driven workflows. In a schematic, a connection is just a net — a logical line that says two things are joined. But in photonics, geometry is loss: every bend, crossing, and detour costs optical power, and a connection that looks trivial on a schematic may be expensive or infeasible to actually route on a chip. By keeping layout, routing, components, hierarchy, and simulation-aware context in one editable place, Qfactr keeps you working against the physics that ultimately decides whether a design works.
Qfactr is not a schematic-only tool, and it is not a pure simulator. It is a representation layer: one model that carries connectivity, true physical layout, hierarchy, and physical constraints together, so the picture you edit is the picture that gets fabricated.
Who it is for
Qfactr is built for the people who design photonic systems and the teams forming around them:
- Photonic chip designers who want layout, routing, and loss in one editable environment instead of stitching tools together by hand.
- Researchers prototyping circuits — filters, interferometers, resonator structures — who need geometry and physics to stay honest while they explore.
- Teams moving from electronic to photonic design, where the schematic-first mental model breaks down because, in photonics, where you place a component determines whether you can even route to it.
You do not need to be a photonics expert to start. If you are new to the domain, Photonic integrated circuits 101 builds the background; if you already design PICs, the Core concepts page maps Qfactr's model to ideas you already know.
What Qfactr does today
These capabilities exist now. Each is covered in more depth elsewhere in the docs.
| Capability | What it means |
|---|---|
| Physical coordinates | Components live in real micrometers — geometry is true from the first placement, not approximated on a grid. |
| Explicit routing | Waveguides are routed as real S-bend and Manhattan paths between component pins, not abstract nets. See Routing waveguides. |
| PDK-aware parts | Component libraries carry physical pin positions and S-matrix (S-parameter) data, so parts behave like the devices they represent. See Process design kits. |
| Hierarchical blocks | Reusable subcircuits with external pins and frozen layouts, composed into larger systems. See Hierarchical blocks. |
| Simulation-aware | Transmission and loss are derived from the actual path geometry; one action surfaces them for the current design. See Simulation-aware design. |
| Open foundation + export | Built on the open Lunima engine, with Nazca export to hand off to an existing fab flow. |
Qfactr also includes an AI assistant that works over the structured circuit state — components, pins, routes, hierarchy, and physical constraints — so natural-language editing stays grounded in the real design. It is not a chatbot bolted onto a canvas; it is covered on the AI assistant page.
How these docs are organised
The documentation is grouped into five sections. Read top to bottom if you are new, or jump straight to what you need.
- Getting started — orientation and a hands-on first build: Your first layout and The workspace.
- PIC design fundamentals — domain background that applies whatever tool you use: photonic circuits, waveguides and loss, components, and PDKs.
- Designing in Qfactr — the product how-tos: placing components, routing, hierarchy, simulation-aware checking, and exporting.
- Physics-informed ML — the direction Qfactr is built around: why photonic layout is hard, the AI assistant, and generative placement.
- Reference — export formats, a glossary, an FAQ, and notes on the open Lunima engine.
Where this is heading
Qfactr's longer-term direction is to bring generative ML to the hardest part of photonic design — arranging components so they can actually be routed at acceptable loss. The aim is not abstract placement but routing-aware layout exploration that respects the physical difficulty of pushing waveguides through dense systems.