Sub-brand · v1

ChipletOS Photonic Signoff

Photonic IC signoff for silicon-photonics designers. Six primitives (waveguide, MZI, MMI, ring, grating, photonic crystal), 5 of 6 on trained AI surrogates with per-prediction 95% confidence intervals, AIM-Photonics-class DRC, and Lumerical / OmniSim / GDSII export — under the same validation discipline as the ChipletOS chiplet stack.

One platform, two markets: chiplet packaging and silicon photonics. Same surrogate stack, same validation suite, same cross-vendor exporter discipline.

Try the live waveguide predictor →Read the architectureStrategic acquirer fit

Six primitives

Waveguide → MZI → MMI → Ring → Grating → Photonic Crystal

Each primitive ships with a fast closed-form analytical hot path for inverse-design loops, plus a trained AI surrogate at ≥ 99% R² vs reference solver for 5 of 6 primitives. Waveguide ships on the analytical model today; an AI-surrogate refresh is on the roadmap.

Waveguide

Live (analytical)

Effective-index model for rectangular SOI strip waveguides + propagation loss + bend loss. Pure-Python at ~0.04 ms / call.

Surrogate target

Closed-form analytical (AI surrogate refresh on the roadmap)

API

POST /v1/photonics/predict-waveguide-mode

Mach-Zehnder Interferometer

Live (AI surrogate)

MZI transmission spectrum from arm-length imbalance + phase shifter + coupler ratios.

Surrogate target

AI surrogate, ≥ 99% R² vs reference solver

API

POST /v1/photonics/waveguide-from-target (chained)

MMI Coupler

Live (AI surrogate)

1×2, 2×2 self-imaging-length MMI power splitter. Length, insertion loss, port imbalance.

Surrogate target

AI surrogate, ≥ 99% R² vs reference solver

API

POST /v1/photonics/predict-mmi (chained)

Ring Resonator

Live (AI surrogate)

Add-drop / all-pass ring transfer function. FSR, FWHM, Q from coupling coefficients + round-trip loss.

Surrogate target

AI surrogate, ≥ 99% R² vs reference solver

API

POST /v1/photonics/ring-from-target

Grating Coupler

Live (AI surrogate)

Fiber-to-chip grating coupler — peak wavelength, 1-dB bandwidth, peak coupling efficiency from pitch + duty + etch depth.

Surrogate target

AI surrogate, ≥ 99% R² vs reference solver

API

POST /v1/photonics/predict-grating (chained)

Photonic Crystal

Live (AI surrogate)

2D triangular-lattice band-gap predictor. Lattice topology surrogate lifts full-wave band-structure runs from hours to milliseconds.

Surrogate target

AI surrogate, ≥ 99% R² vs reference solver

API

POST /v1/photonics/predict-photonic-crystal (chained)

Live API endpoints

11 curated /v1/photonics/* endpoints

Three public + eight auth-gated routes. Try the public ones below; auth-gated ones are listed for completeness — request access via the buyer app.

GET/v1/photonics/signoff-healthPublic

Returns availability of the analytical hot path, full-wave reference solver, validation-suite status, and surrogate-version metadata.

POST/v1/photonics/drc-photonicPublic

AIM-Photonics-class DRC: minimum waveguide width, bend-radius vs material, ring gap, grating pitch. Returns violation list keyed to AIM rule IDs.

{"geometries":[{"type":"waveguide","width_nm":50,"bend_radius_um":10,"material":"Si"}]}
POST/v1/photonics/validate-against-ieeePublic

Runs the analytical photonic stack against an IEEE literature corpus (SOI Neff papers) and returns per-paper MAE + a pooled verdict.

{}
POST/v1/photonics/predict-waveguide-modePublic

Returns effective index Neff, group index ng, propagation loss (dB/cm), and bend loss (dB/turn) for a rectangular SOI strip waveguide.

{"width_nm":450,"height_nm":220,"wavelength_nm":1550,"material":"Si"}
POST/v1/photonics/waveguide-from-targetAuth-gated

Inverse design: given target Neff, return manufacturable (width_nm, height_nm) and back-checked Neff.

{"target_neff":2.4,"wavelength_nm":1550}
POST/v1/photonics/ring-from-targetAuth-gated

Given target FSR and Q, return ring radius + coupling coefficient + predicted ER.

{"target_fsr_ghz":100,"target_q":10000}
POST/v1/photonics/resonator-paretoAuth-gated

Multi-objective Pareto front over ring resonator design space.

{"objectives":["fsr","q","footprint"]}
POST/v1/photonics/waveguide-paretoAuth-gated

Multi-objective Pareto front over waveguide design space.

{"objectives":["loss","footprint","neff_band"]}
POST/v1/photonics/export-lumericalAuth-gated

Export a photonic design to a Lumerical-importable .lsf / .ldf bundle. Bit-perfect native validation runs customer-side in the licensed Lumerical instance.

{"design_id":"..."}
POST/v1/photonics/export-omnisimAuth-gated

Export to Photon Design OmniSim format. Bit-perfect native validation runs customer-side in the licensed OmniSim instance.

{"design_id":"..."}
POST/v1/photonics/export-gdsiiAuth-gated

Export a closed-loop-verified photonic layout to GDSII for any AIM-Photonics-class foundry.

{"design_id":"..."}

Solver stack

Reference FDTD → analytical models → trained AI surrogate

Native solvers never belong in an optimizer inner loop — a hard-won lesson from our chiplet stack that the photonic sub-brand inherits on day one. The reference FDTD solver runs in isolation; the fast analytical models and trained AI surrogates do all the hot-path work.

Reference truth

Full-wave FDTD reference solver

Process-isolated for stability. Generates training corpora and verifies the surrogate before promotion. Never called in an optimizer inner loop.

Cross-physics verification

Independent FDTD

Independent solver implementation for the cross-solver agreement check. Currently on the roadmap.

Fast analytical hot path

Closed-form analytical models

Pure-Python effective-index + transfer-matrix + coupled-mode models. ~0.04 ms / call. The hot path for inverse-design loops.

AI surrogates (v1 live)

Trained MLP surrogates with per-prediction CI

5 of 6 primitives ship with a trained AI surrogate at ≥ 99% R² vs reference and a 95% confidence interval on every prediction. Waveguide ships on the analytical model with an AI-surrogate refresh on the roadmap.

Scope

What is live today and what is on the roadmap

Live today

  • 5 of 6 primitives on trained AI surrogates at ≥ 99% R² vs reference solver
  • Per-prediction 95% confidence interval on every surrogate call
  • Waveguide-mode predictor on the closed-form analytical model
  • AIM-Photonics-class design-rule check (8 rule families)
  • Published-paper cross-check against 5 silicon-photonic references
  • Eight auth-gated routes: inverse design, Pareto, 5 cross-vendor exporters

On the roadmap

  • AI-surrogate refresh on the waveguide primitive (currently on the analytical model)
  • Higher-fidelity surrogate retrain against full reference FDTD ground truth
  • Independent cross-solver agreement check (Meep vs gprMax disagreement matrix)
  • Vendor-native bit-perfect validation in licensed Lumerical / OmniSim / ANSYS / KLA instances
  • Photonic wet-lab VNA campaign for measured-truth verification (analogous to our chiplet wet-lab pack)

See Trust & Validation for the full validation-suite breakdown.

For strategic acquirers

Widens the acquirer pool 2×

Chiplet alone narrows the natural acquirer set to Cadence, Synopsys, ANSYS, Intel Foundry, and Corning. Adding the photonic sub-brand doubles it: Synopsys-Lumerical, ANSYS-Lumerical-adjacent, Photon Design, and the captive Marvell SiPh / Acacia (Cisco) / Lumentum design teams all become natural strategic fits.

Synopsys (Lumerical)

Owns Lumerical (~$80M photonic EDA revenue). ChipletOS Photonic Signoff is the AI-native distribution layer Lumerical doesn't have.

View buyer page →

ANSYS (Lumerical FDTD)

No AI-EDA story, no photonic enterprise pull-through. ChipletOS photonic + chiplet stack is the build-vs-buy decision.

View buyer page →

Photon Design (Omnisim)

Smaller incumbent; natural partnership candidate for joint distribution under the ChipletOS Photonic Signoff brand.

View buyer page →

Try the live waveguide predictor

Three public endpoints, no auth, real Modal compute. The same audit pattern as the chiplet stack — applied to silicon photonics.

Open the photonic playground →Read the architecture