Nowadays, wearable devices and sensor nodes for IoT applications pervade our lives.
Integration of machine intelligence in these devices would allow fast decisions based on data collected from their sensors, with increased data security. In these battery-powered devices energy efficiency is key.
AI accelerators based on Analog in-Memory Computing (AiMC) will drastically reduce dynamic power consumption, as they reduce data transfers from and to the memory, and parallel analog computation is much more energy efficient than digital computation for the precision required in properly optimized Deep Neural Networks. In systems with low duty cycles such as wearables and sensor nodes, static and standby power of the AiMC macro can significantly limit the AI integration.
In this project, you will investigate in detail the static power consumption in AiMC macros and develop and implement novel design solutions to push more AI towards edge devices.
Percentage of work
50% design and simulation`
Type of project: Internship, Thesis, Combination of internship and thesis
Required degree: Master of Engineering Technology
Required background: Electrotechnics/Electrical Engineering