Cogitan

Products

Applied AI.
Built to ship.

Each product starts with a well-defined engineering problem, a simulation or verification bottleneck, and a clear path to measurable speedup with calibrated confidence.

Active

01

Fluxus

Active · SFQ5ee+

Neural design verification for superconducting RSFQ circuits

A learned physics model trained on your process design kit. Given any RSFQ cell layout, it predicts the outputs that matter — timing, current, and design rule compliance — with a confidence bound on every result, in milliseconds per cell.

The model is PDK-agnostic. We train a custom surrogate on each client's cell library and process node. Proof of concept is complete on MIT-LL SFQ5ee+. Additional process nodes are onboarding now.

Propagation delayPeak currentDRV detectionFunctional classification1σ confidence boundsPDK-agnosticDifferentiable
Request access

Key results

Speed vs. SPICE
200–500×

chip-scale

Functional accuracy
100%

validated types

Delay prediction
sub-3 ps

SPLIT/DFF/AND2

DRV detection
99.2%
Uncertainty
1σ bounds

every output

SPICE fallback
Auto-flag

low-confidence

Compatible processes

MIT-LLIPHTSeeQCAISTSkyWaterCustom

In development

02

New product coming

03

New product coming

Your domain?

If you have a simulation bottleneck in a physics-constrained engineering domain, we want to hear about it.

Get in touch