Collaborating with ITB, Nodeflux Develops "Deep Learning" Capable Chip

Nodeflux, a startup developer of AI-based technology solutions, announced its collaboration with the Bandung Institute of Technology (ITB) for the IC Lab Design section. This collaboration is the first step for both parties to conduct research and development hardware Deep Learning Inference Chip (DLIC).

It was explained that this DLIC is designed to have an architecture deep learning implemented at the hardware level. It is hoped that the computing process can be faster with lower power consumption. The algorithm that will be implemented in a specific IC developed by Nodeflux.

DLIC is one of the basic components of computer equipment in the form of an integrated circuit that works using a series of algorithms deep learning through visual data in the form of photos, images, and videos. Tools can generate structured data and produce valuable information as stakeholder material in making strategic decisions.

This collaborative project will run for approximately 12 months through the gradual development of DLIC. Co-founder & CTO nodeflux Faris Rahman explained that the development of this DLIC will take place in several stages. For the initial stage like now, what is being done is architectural design, both in terms of the algorithms implemented and also the components hardware to be able to abstract the algorithm.

The next is synthetizing and Development. The phase to determine the parameters of the proposal to be achieved, such as computational operating capacity per unit time, computational speed, and computational accuracy performance.

"Next is testing and verification to ensure that all reference parameters are met. In this collaboration, we will try to see the performance of the architectural design that will be prototyped, in the form of a Single Board Computer. Furthermore, if we have verified the capabilities of the architectural design, we will carry out further cooperation to be able to enter the fabrication design stage," explained Faris.

Currently, the use of DLIC is still limited for solutions related to Nodeflux products in Vision AI through VisionAire. However, it is possible to develop it en masse, but there is still a long way to go considering this is only in its early stages.