Novel Memristive Technology for AI-Centric Memory and Computing
Project ID: 2531bd1665
(You will need this ID for your application)
Research Theme: Information and Communication Technologies
Research Area(s):
Artificial intelligence technologies
Microelectronic device technology
UCL Lead department: Electronic and Electrical Engineering (EEE)
Lead Supervisor: Adnan Mehonic
Project Summary:
Why this research is exciting and important
Modern computing is increasingly limited by the speed, cost, and energy demands of memory. This challenge is particularly acute for artificial intelligence, where most of the energy consumed by large models is spent moving and storing data rather than performing computation. This project explores an emerging and transformative concept known as Managed Retention Memory, recently proposed by Microsoft, which redefines how memory systems store information by allowing data to be retained only as long as it is useful, leading to dramatic improvements in energy efficiency, performance, and scalability for future AI and data-centric computing. Implemented using memristive devices such as Resistive RAM (ReRAM), the project aims to create adaptive, high-performance memory that combines efficiency, scalability, and long-term reliability, opening new directions for sustainable AI hardware.
Who you will be working with
You will join UCL’s Department of Electronic & Electrical Engineering, supervised by Dr Adnan Mehonic, and collaborate with Intrinsic Semiconductor Technologies Ltd, a UCL spin-out developing next-generation memristive technology. There are plenty of opportunities to engage with a wider network of industrial collaborators throughout the project.
What you will be doing
- Design, fabricate and test memristive devices.
- Learn about the physical mechanisms that govern device operation and explore design improvements to enhance performance and reliability.
- Design and simulate circuits and architectures to explore and benchmark key design choices, identifying optimal trade-offs.
- Benchmark performance, energy efficiency, and reliability in realistic AI inference and caching workloads.
Who we are looking for
We welcome applicants with an undergraduate or Master’s degree (minimum upper second class, 2:1 or equivalent) in Electronic Engineering, Physics, Materials Science, or Computer Science. You should be curious, creative, and motivated to work across materials, circuits, and AI hardware in a dynamic and fast-evolving research environment.