Energy and Cost Efficient AI

Purpose and goal

The goal of the project is to develop a prototype of a product where user can train AI models based on their proprietary data, especially over large scale data. We will explore the business possibilities of licensing our software for consumer electronics vendors and test the algorithms in the vendor’s simulation environments, with initial focus on 5G/6G applications.

Expected effects and result

This project starts at TRL level _3__ and is intended to end at TRL level _5__. As we have described above, we start at TRL 3, where we have experimental proof of the concept on small and medium scale. The prototype will enable us to validate the performance of our technology in relevant industrial environments. In our case, we will achieve TRL 5 by developing our prototype and testing our technology in telecom vendors’ simulations environments.

Planned approach and implementation

We will start by algorithm testing and verification on customers data, especially for vendors of user electronic devices, telecom vendors, and IoT devices. We will also implement our tools as a PyTorch library. We will spend 6 man months of algorithm validation and testing. The second WP considers development of PyTorch libraries and will take 6 man months. The third work package focuses on product evelopment together with potential users and we will iterate product designs to find a prototype with as user-friendly interface as possible. This will take about 12 man-months.