Electronic-photonic Architectures for Brain-inspired Computing

feed news

Highlighting women’s contribution to the HYBRAIN project – Interview with Elena Ferro

HYBRAIN - Elena Ferro interview

Interview with Elena Ferro, IBM Zurich.

Can you talk a little bit about your work as a researcher in your organisation?

I am in my second year of PhD. My current research focuses on the hardware design of digital near-memory peripheral components for Analog-in-Memory computing, aiming to maximise throughput and energy efficiency. The objective is to fully leverage the potential benefits of In-Memory computing technology while optimizing its performance parameters.

Can you share with us the story/reason that led you to decide to become a researcher?

Upon completing my master’s degree in Nanotechnologies for Information and Communication Technologies, I wanted to increase my expertise in the field of Electrical Engineering. Pursuing a PhD was a good compromise to achieve my objective while still being in contact with the academic community and with the quest towards resolving unanswered questions.

What are the domains that your research is focusing on?

AI, Edge Computing, Electrical Engineering, Digital hardware design and In-memory computing.

Why did you decide to focus on these domains, what do you like about them?

I have always had a strong interest in computer programming and a curiosity about how computers work. As artificial intelligence has become increasingly prominent, I find it fascinating to explore ways to enhance hardware capabilities to better support AI applications.

About your work in HYBRAIN

Why did you decide to work on HYBRAIN?

My main focus at IBM is to improve the Digital periphery of Analog in-memory computing systems and I personally enjoy the idea of exploring novel architectures and methodologies. The HYBRAIN technology, with its unconventional and innovative computing approach, presents an intellectually stimulating opportunity for researchers and engineers, such as myself.

What are you working on in HYBRAIN?

I am working on the digital periphery of near-in-memory computing systems.

Can you provide more details on these activities for our readers, what is it all about?

Since Analog-in-memory computing suffers from some non-idealities of the systems, the precision that is given at the output is not accurate. For this reason, it is common to adopt some digital post-processing that performs what is called “Affine Correction” to improve the accuracy of such systems. However, the most common way of implementing it is to adopt “Floating Point 16b” precision, which implies high precision at the output at the cost of a significant area overhead and power consumption and, therefore, a significant latency to perform all the computations. I am currently focusing on improving such an interface to achieve a precision similar to “Floating Point 16b”, but requiring less area and high parallelisation capacity, i.e. lower latency.

What is your specific personal contribution to these activities?

I am involved in the design of the near-memory digital logic of analog-in-memory computing. My main contribution is to design digital components, simulate and synthesise them.

What is the expected outcome/impact of your work?

My work will contribute to the possibility of using analog-in-memory computing systems in an increasing number of fields where high throughput, high accuracy, and reduced area are needed.

What do you like about the HYBRAIN project?

The HYBRAIN vision. It aims at adopting/exploring brain-inspired hybrid architecture to significantly reduce power consumption and improve computational efficiency in various applications, including AI and IoT. This means that our work in this field can contribute to solving some of the world’s most pressing challenges while driving innovation in the broader tech industry.

More specifically, what do you like about working on HYBRAIN?

One of the most exciting aspects of working on HYBRAIN is the opportunity to push the boundaries of computing technology by adopting analog in-memory computing accelerators. Analog accelerators represent a fascinating intersection of hardware and software, and they have the potential to revolutionise the way we perform certain types of computations. Furthermore, this domain offers a unique blend of creativity and engineering. Designing and optimising analog compute accelerators requires innovative thinking to develop novel architectures and circuit designs. It is a field that allows me to learn and adapt as technology evolves continuously.

Your personal insights

The HYBRAIN technology will enable various innovative AI/Edge Computing applications across the health, automotive, cybersecurity, climate change domains and more. Personally, what is the potential application for the HYBRAIN technology that excites you the most?

The potential application for HYBRAIN technology that excites me the most is its role in advancing healthcare and improving patient outcomes. The healthcare sector increasingly relies on AI and edge computing to enhance diagnostics, treatment, and patient care. HYBRAIN’s capabilities in edge computing can be employed in the context of real-time diagnostics to develop real-time diagnostic tools. For example, the HYBRAIN technology can be used for wearable devices or sensors that continuously monitor vital signs, analyse data locally, and provide immediate feedback to healthcare professionals or patients. This could lead to earlier detection of health issues, more timely interventions, and ultimately save lives.

In your opinion, why should AI and Edge Computing researchers pay attention to the HYBRAIN technology?

HYBRAIN represents an innovative approach to AI and Edge Computing. It combines the strengths of analog computing with digital processing, offering researchers a new computing paradigm to explore. Its technology offers a unique combination of efficiency, low latency, scalability, and customisation, all of which have the potential to advance the state-of-the-art in edge-based AI applications across a wide range of industries.

In your opinion, why should AI and Edge Computing organisations pay attention to the HYBRAIN technology?

The primary objective of HYBRAIN is to improve the energy efficiency, latency, and power consumption of hardware systems. Therefore, organisations focusing their attention and integrating the HYBRAIN technology can leverage these enhancements to position themselves for success in the evolving landscape of edge-based AI and computing.

If HYBRAIN is successful in creating its technology, how do you think the AI and Edge Computing domains will change as a result of the uptake of HYBRAIN’s technology?

The integration of HYBRAIN’s technology is expected to lead to a substantial increase in the efficiency of AI and Edge Computing systems. This would result in reduced power consumption, longer device battery life, and overall improved energy efficiency, making edge devices more sustainable and cost-effective. Furthermore, HYBRAIN’s local processing capabilities could dramatically reduce latency in edge computing applications. Ultra-low latency processing is critical for real-time decision-making in autonomous vehicles, robotics, and augmented reality, enhancing user experiences and safety. In my opinion, it will be a step forward for the current computing paradigm.

What would be your advice for young women wishing to pursue their passion for AI, Edge Computing and other STEM domains with a research career?

My suggestion for young women, which can be applied to young men as well, is to define your career goals and the areas of research that genuinely excite you. Having a clear vision will help you stay motivated and focused in pursuing your goal.