Mission

Light does the math.

Photonic chips already power optical communications, sensing, quantum systems and emerging compute engines. Designing them is still stuck in the past: fragmented tools, manual workflows, no AI assistance. We think that is a solvable problem, and worth solving properly.

Vision

A world where photonic chips are designed as easily as digital ones.

One platform that designs any photonic chip — from an optical transceiver to a quantum sensor to a compute engine. Engineers describe what the chip should do, and the platform carries it from there.

Mission

Collapse the photonic design cycle from months to days.

Build the AI-orchestrated platform that designs photonic chips — where AI agents design, verify and package, and deterministic engines prove the result. Whatever the chip computes, carries or senses, it is designed in one place.

The problem

Designing a photonic chip today is a nightmare of disconnected tools

A single photonic chip needs expertise across five domains at once, each with its own tools, languages and workflows — and every decision couples to every other.

The five domains and the tools each one usually drags in
Domain	What is needed	Usually spread across
Photonic design	Rings, modulators, detectors, couplers, waveguides	Photonic layout, mode, FDTD and BPM tools
Electronic design	Driver ASICs, DAC arrays, analogue cells	Schematic capture, SPICE, place-and-route
Firmware	Sequencers, calibration, control logic	RTL tools, FPGA vendor flows, compilers
Carrier board	Wire-bond, power delivery, high-speed routing	PCB layout, SI and PI analysis
Advanced packaging	Interposer, thermal, test	No unified tool; spreadsheets and hand-off

Months per cycle

Most of the time goes into translating between domains, not into designing in any one of them.

Re-spins found late

Thermal, optical and electrical constraints collide after tapeout, when a fix is a new mask set.

Architectures never tried

Each promising direction costs compute and engineering weeks, so most of the design space is never evaluated at all.

Commitments

What constrains how we build this

These are not aspirations. They are the rules that decide what we are allowed to ship and what we are allowed to claim.

No fabricated results

Every number we publish is backed by a reproducible run. We do not inflate, and we do not round in our own favour.

Transparent limitations

We state what is simulated versus measured, and what is validated versus projected. Where a process or a method has gaps, we name them.

Responsible AI

AI agents propose; deterministic engines dispose. There is no AI sign-off without formal verification behind it, and no agent approves its own work.

IP respect

We honour foundry NDAs, third-party licences and open-source contributions — and we expect the same of our supply chain.

Safety-conscious hardware

ESD, reliability and test coverage are first-class parts of the flow, not checkboxes added before a review.

Provenance

Every artifact carries the tool, the inputs and the status that produced it, so a claim can be traced back to the run behind it.

Who we serve

Two sides of the same problem

Design teams building photonic chips

Optical transceivers & CPO — photonic design, link budget, thermal and packaging in one flow.
Photonic compute — full-stack design from component to packaged chip.
Quantum sensing — ring-based sensors with calibration and verification that can be repeated.
Defence, aerospace & space — full design control, deterministic verification, assured timing.
University research — accessible tools for novel rings, modulators and couplers.

Supply-chain partners

Foundries — we make your process design kit design-ready. More designers succeed on your process, on purpose rather than by luck.
Component manufacturers — we make your parts design-integrated, so a designer can find, simulate and specify them in the same place they design.
Test & assembly — production test stages and capacity are modelled in the flow, so a yield decision is made on the whole chain.

Partner with us →

Why Lumicule

What changes, and what does not

The workflow does not get less rigorous. It gets faster where it was slow.

The traditional workflow against ours
The traditional workflow	With Lumicule
Five disconnected toolchains	One integrated platform
A handful of options evaluated per cycle	Thousands of candidates explored, bad ones killed early with evidence
Cross-domain conflicts discovered late	Co-design across domains from day one
Verification as an afterthought	Deterministic verification in every stage — temporal, vacuity-gated, agent-actionable
Proofs that do not connect to the design	Faithfulness checks before any formal result is trusted
Designer judgment limited by experiment budget	Designer judgment applied to far more evidence
Foundry integration is painful	A foundry-ready flow with process kits integrated
No AI assistance	Specialized agents at every step — proposing, never signing off
Debug = “make the test pass”	Debug = “root cause found”, through the RCA toolchain

Work with us

If photonics is on your critical path, we should talk

Tell us what you are building, and we will tell you honestly where the platform helps today.

Open the workbench Talk to Lumicule