Nikhil Bhendawade

Title: Nikhil Bhendawade and Resource-Efficient Neural Network Attention

Introduction

Nikhil Bhendawade is an inventor based in Cupertino, California, whose patent work addresses efficient attention mechanisms in neural networks. His listed patent focuses on reducing the processing and memory demands associated with attention.

Latest Patents

Bhendawade is associated with one patent, titled “Resource-efficient attention in a neural network.” The invention describes an attention mechanism that generates attention information using head-specific query information alongside shared key and value information. It avoids computing and caching separate, head-specific key and value information, helping conserve processing and memory resources.

The approach can be used in the decoder of an encoder-decoder system or in a standalone decoder. In some implementations, the attention information supports generating synthesized text from input text.

Career Highlights

Bhendawade is associated with Microsoft Technology Licensing, LLC. His listed patent reflects work on improving the resource efficiency of neural network systems.

Collaborations

The provided information lists Yu Yan and Jiusheng Chen as Bhendawade’s coworkers. Their names are associated with his professional context, though specific collaborative roles in the patent are not detailed here.

Conclusion

Nikhil Bhendawade’s patent describes a neural network attention approach designed to reduce computation and memory use while supporting text generation.