Differentiable Silicon
Tesseract Hackathon 2026

Case study

Research / Differentiable TCAD

Open Source

Dhaka, Bangladesh

Research / Differentiable TCAD · Detailed breakdown

Project overview

Why this build
mattered.

One gradient from an ECG classification loss, through a PyTorch spiking network and a JAX device model, into Synopsys Sentaurus TCAD. Best-in-Track winner, Hybrid ML and Mechanistic Models track.

Snapshot

Tesseract Hackathon 2026 / Open Source / Python, PyTorch, JAX

PythonPyTorchJAXTesseractSentaurus TCADDEVSIM
1.400 to 1.018
loss
0.250 to 0.688
accuracy
64 calls
solverCalls

System views

Screens and
surfaces

What shipped inside

Key features
and modules

The architecture, workflow and results that made this project useful.

Highlight

Sentaurus is a closed binary with no adjoint, so the gradient comes from trust-region finite differences plus a Broyden update

Highlight

Balanced cross-entropy 1.400 to 1.018 and accuracy 0.250 to 0.688 in 64 solver calls

Highlight

An open DEVSIM backend can replace Sentaurus without changing the pipeline

Delivery summary

Build notes

Project essentials

Differentiable Silicon

Tesseract Hackathon 2026 · Open Source · Research / Differentiable TCAD

Sentaurus is a closed binary with no adjoint, so the gradient comes from trust-region finite differences plus a Broyden update1
Balanced cross-entropy 1.400 to 1.018 and accuracy 0.250 to 0.688 in 64 solver calls2
An open DEVSIM backend can replace Sentaurus without changing the pipeline3

Let's build something worth building.