What is a device model?

  In the semiconductor chip fabrication process, device model is the “bridge” between the fabricated device and IC design. Smaller, faster transistors that consume less power are continuously developed in the fabrication facility, but the final impact that they have can only be evaluated after being evaluated inside a circuit. Device models enable this circuit evaluation, containing the relevant electrical information of the transistor. Thus the speed and accuracy of the device model determines the speed and accuracy of the circuit-level prediction.

Problem

  Today, device models are formula-based models derived by humans, which means:

① Need for semiconductor expert intervention

② Long development period (years)

③ Limited accuracy

Our Solution

Alsemy presents a new paradigm

in which device models are data-based models generated by deep learning, which enables:

① No need for expert intervention

② Short development period (hours)

③ Improved accuracy

Distinctive Features

​

1. Ease-of-use by anyone

Alsemy offers an easy-to-use AutoML algorithm that even non-AI experts can use.

Based on customer's data, the algorithm automatically determines the optimal network structure.

​

2. Less data-size dependence

In order to cope with situations where data is limited,

Alsemy uses knowledge transfer algorithms to incorporate good prior knowledge

in learning process.

​

​

3. Device physics based

Purely data based ML models may violate semiconductor device physics. Alsemy’s physics-informed AI model can resolve this issue.

Vision

Bringing better semiconductor technology to the world faster

by full automation of chip design and optimization

Contact Us