Shalman Jesse Ngozi Ojukwu

• MEng (Hons) Electronics and Computer Engineering with Industrial Project
• Designing Pervasive AI Hardware using Tsetlin Machines

In future, ubiquitous applications will need Machine Learning everywhere for artificial and augmented intelligence – from sensed data at the micro-edge to the decision processes in the cloud. Two major challenges of these applications are energy efficiency and unreliability (i.e. high variability) of power supply from batteryless harvesters. The current generation of AI hardware in the form of Deep Neural Networks are highly energy consuming and not suitable for implementation under power supplies that vary significantly over time.

How will AI hardware look like for these applications? How can we design hardware systems that can autonomously adapt to the natural power variations coming from harvesting sources? These questions will critically shape the way we design new-generation hardware circuits and systems for pervasive applications.

Tsetlin Machine has recently emerged with improvement over traditional Machine Learning in terms of faster convergence and minimum training time. Dr Shafik’s group has recently been investigating ways to incorporate energy autonomy and efficiency features for this new Machine Learning hardware by using its natural capability to alter complexities under extreme power situations with minimal loss of accuracy. The group has a number of collaboration initiatives with national and international academic and industrial research groups.

Funding source: Newcastle University

Project supervisor: Dr Rishad Shafik