SPINS Photonics

HyperWave

Run FDTD simulations and inverse design optimization on NVIDIA B200 GPUs. Python SDK, Jupyter notebooks, pay-per-use.

Get StartedRead the Docs

Up and running in minutes

Three steps from install to GPU-accelerated simulation results.

01

Install

pip install hyperwave-community

Install the Python SDK with a single command. Works in local environments, Jupyter notebooks, and Google Colab.

02

Configure

hwc.build_recipe(component_name=...)

Start from GDSFactory components or define arbitrary device geometries. Configure materials, resolution, and simulation parameters in Python.

03

Run

hwc.run_simulation(recipe_result=...)

Execute on B200 GPUs in the cloud. Get results in minutes, not hours.

api_workflow.py
pip install hyperwave-community

import hyperwave_community as hwc
hwc.configure_api(api_key="your-api-key")

# Build device from GDSFactory component
recipe = hwc.build_recipe(
    component_name="mmi2x2_with_sbend",
    resolution_nm=20,
    n_core=3.48,
)

# Build monitors and solve waveguide mode
monitors = hwc.build_monitors(recipe, source_port="o1")
source = hwc.solve_mode_source(recipe, monitors)

# Run on B200 GPU
results = hwc.run_simulation(recipe, monitors, source)

# Analyze transmission
hwc.analyze_transmission(results)

Unparalleled Developer Experience

A clean, Pythonic API designed for photonics researchers. Define your simulation, launch it, and get results, all from a notebook.

  • B200 GPU Backend

    192GB HBM3e with native CUDA compilation for maximum throughput.

  • Automatic Adjoint Differentiation

    Calculate gradients for inverse design with zero overhead.

  • Jupyter-Native Workflow

    Inline plots and seamless Jupyter and Colab notebook integration.

  • GDSII Import and Export

    Build from GDSFactory components and export fabrication-ready designs.

Inverse Design in Action

Watch gradient-based optimization converge to high-efficiency photonic devices in real time.

Simple, Transparent Pricing

No tiers. No hidden fees. Just compute.

Individual

$25/hour of GPU simulation
No subscriptions or commitments
Credits never expire
Pay only for compute time
B200 GPU with 192GB HBM3e

Enterprise

Custom Pricing

On-premises deployment support
Custom GPU infrastructure
Dedicated engineering support
Volume licensing available

Why HyperWave

Purpose-built for photonics researchers who need speed, precision, and a modern developer workflow.

Accelerate Design Cycles

NVIDIA B200 GPUs with 192GB HBM3e deliver simulation results in minutes. Run high-resolution 3D FDTD simulations that would take days on CPU solvers like Lumerical or COMSOL.

Purpose-Built for Inverse Design

Automatic differentiation enables gradient-based topology optimization that converges to fabrication-ready designs.

Python-First Developer Experience

A clean Python SDK with Jupyter and Colab support, letting you offload heavy compute to cloud GPUs. GDSII import/export included. Script your entire workflow and version control your simulations.

Ready to accelerate your photonics design?

Get started with free credits and run your first GPU-accelerated simulation today.

Create AccountView Documentation