Gurshaant Singh Malik

Title: Gurshaant Singh Malik and Low-Power Recurrent Neural Network Hardware

Introduction

Gurshaant Singh Malik is an inventor based in Waterloo, Canada, associated with Applied Brain Research Inc. His patent work focuses on hardware architectures for efficiently implementing recurrent neural networks, particularly in applications where power consumption matters.

Latest Patents

Malik is credited with 1 patent, titled “Partitionable Digital Hardware System for Implementing Recurrent Neural Networks.” The invention maps recurrent neural networks to hardware computation blocks designed for Legendre Memory Unit (LMU), Projected LSTM, and feed-forward cells. It uses iterative resource-allocation algorithms to partition networks and time-multiplex them across distributed computation blocks, with optimization for power, performance, and accuracy.

The system is intended to support low-power, high-performance deployment in battery-sensitive applications, including automatic speech recognition, keyword spotting, biomedical signal processing, and other time-series processing tasks.

Career Highlights

Malik’s patent reflects work at the intersection of neural-network computation and specialized digital hardware. By adapting computation to different recurrent-network cell types and balancing resource use against performance and accuracy, the invention addresses practical constraints in deploying machine-learning systems on power-limited devices.

Collaborations

Malik is associated with Applied Brain Research Inc. His listed coworkers include Christopher David Eliasmith and Aaron Russell Voelker.

Conclusion

Gurshaant Singh Malik’s patent highlights an approach to making recurrent neural networks more practical for energy-conscious applications through partitionable, optimized hardware.