Quantum Surrogate Tools — v0.1

Engineering
discovery
at qubit scale.

Ansatz maps the design space of qubit sensing systems using physics-informed surrogate models — so you spend fewer cycles on the bench and more time understanding why something works.

Designed for precision measurement teams

ansatz / qubit-array-viz
live
97.4%
Avg. Fidelity
3×3
Array Size
16ms
Sweep Rate

Integral design

Physics-informed, not curve-fit

δ

Sensitivity first

Know what matters before you build

σ

Uncertainty-aware

Every prediction ships with confidence bounds

Tensor-native

Designed for multi-qubit coupling from day one

The method

Structured discovery,
not guesswork.

Every step in the Ansatz workflow is designed to generate trustworthy, reproducible insight — not just a faster path to a number you can't explain.

Step 01

Define your sensing objective

Specify qubit type, target frequency, coupling topology, and the performance metric you care about — sensitivity, coherence time, or noise floor.

Specification
Step 02

Run the surrogate sweep

Ansatz trains a physics-informed surrogate model on your design space, sampling intelligently to cover the parameter manifold without exhaustive simulation.

Exploration
Step 03

Inspect the landscape

Visualize sensitivity maps, identify promising operating points, and understand which parameters dominate performance — before committing to fabrication.

Analysis
Step 04

Validate against experiment

Upload measurement data from your lab. Ansatz compares predictions to reality and refines the model — closing the loop between simulation and bench.

Validation

Capabilities

Built for the real
complexity of qubits.

Qubit sensing design involves dozens of coupled parameters across material, geometry, and control. Ansatz is built for that complexity from the start.

Surrogate modeling

Physics-informed Gaussian process surrogates trained on sparse simulation data — accurate interpolation across the full design manifold.

Gradient-based sensitivity

Automatic sensitivity analysis reveals which parameters drive performance, so you focus fabrication effort where it counts.

Multi-objective Pareto fronts

Jointly optimize sensitivity and coherence. Ansatz surfaces the trade-off frontier so you choose your operating point consciously.

Structured parameter sweeps

Intelligent Latin hypercube and Sobol sampling covers the space efficiently — fewer runs, more coverage, reproducible results.

Coupling-aware layout tools

Model qubit–qubit coupling, cross-talk, and substrate modes as first-class design variables, not afterthoughts.

Experiment-in-the-loop

Upload lab measurements. The model refines itself against reality, narrowing prediction uncertainty as your data accumulates.

Validation

Claims backed by
measurement.

We don't publish benchmarks we can't reproduce. These numbers come from internal validation on superconducting transmon and flux qubit designs.

10×
Fewer simulation runs
vs. grid search on a 6-parameter qubit design
97.4%
Surrogate accuracy
Mean R² on held-out test designs
<5%
Prediction error
On sensitivity coefficient vs. full FEM
3–5×
Faster design iteration
Reported by early-access teams
ansatz run --validate qubit_design_v3.yaml
09:14:02surrogate.fit(): 128 training points loaded
09:14:04GP kernel: Matérn 5/2 + RBF (ℓ=0.34)
09:14:09cross_val R²: 0.974 (10-fold)
09:14:12sensitivity.compute(): ∂S₂₁/∂d_gap dominant
09:14:15pareto.sweep(): 840 candidates evaluated
09:14:18front_size: 12 non-dominated designs found
09:14:20export: JSON + SVG layout ready
09:14:21

* Results on internal benchmark datasets. Performance varies by design complexity and parameter dimensionality.

Early access

Start with the qubit
sensing toolkit.

Ansatz is in active development. The qubit sensing design module is available now for research teams. We're expanding deliberately — building the right tools for the right problems.

Superconducting transmon support
Flux qubit coupling models
Sensitivity sweep reports
Experiment data import
Open Ansatz

No credit card required for research teams