Laith A Shamieh

Title: Laith A. Shamieh and Analog Compute-in-Memory Innovation.

Introduction

Laith A. Shamieh is an inventor based in Charlotte, North Carolina. His patent work includes an architecture designed to make neural-network computations more efficient by keeping signals in analog form as they move between processing stages.

Latest Patents

Shamieh is associated with 1 patent, titled “Fully Analog Compute-in-Memory Architecture for Neural Networks.” The invention describes a Fully Analog State-Space Compute-In-Memory (FAST-CIM) architecture that connects cascaded compute-in-memory arrays while maintaining continuous analog signal flow.

In the proposed design, a feedforward array performs an initial vector-matrix multiplication using computations within memory cells. Its analog output passes through a gain circuit, which converts current signals to voltage signals, and then flows directly to a recurrent array without digital conversion. The recurrent array performs State-Space Model computations using the incoming signals and previous state information stored by a State Write and Propagate (SWAP) circuit with capacitive elements.

By avoiding analog-to-digital and digital-to-analog converters between processing stages, the architecture aims to reduce power consumption and latency compared with implementations that rely on digital interfaces between arrays.

Career Highlights

Shamieh is associated with Georgia Tech Research Corporation. His patent reflects work at the intersection of analog computing, memory architecture, and neural-network processing.

Collaborations

Shamieh’s listed coworker is Saibal Mukhopadhyay. Their association connects the invention with research into compute-in-memory systems and efficient neural-network hardware.

Conclusion

Laith A. Shamieh’s patent presents a fully analog approach to linking compute-in-memory arrays for neural-network computations, with the goal of reducing conversion overhead, power use, and latency.