PC : PC: Could you tell us about your background and how you came to work on these complex topics of computing and reliability in the cloud?

AE : I was born in Canada, where I completed my engineering studies. I had the opportunity to do an exchange program at UTC Compiègne and then pursued a DEA (master’s degree) at Jussieu in Paris. Afterwards, I returned to Canada and worked in networking, notably at Cisco Systems. There, I designed integrated circuits for large routers commonly found in cloud infrastructures. Some are about the size of a pizza box, while others are as big as a fridge!

In this field, reliability is a key challenge. We often talk about throughput, latency, or quality of service, but we tend to overlook that these devices must remain operational despite very rare events, such as errors caused by cosmic radiation. Even at the Earth’s surface, these particles can disrupt the functioning of chips. For a data center with thousands of servers, these errors become noticeable… and thus problematic.

This awareness led me to focus on reliability. I moved to France ten years ago for family reasons and started a PhD at the age of 40 — a bit crazy, maybe, but it was a genuine passion. Since then, I have continued down this path as a research engineer at CEA in Grenoble, where my work focuses on 3D circuits, heterogeneous architectures, and computing reliability.

PC : What sparked your particular interest in cloud computing and digital architectures?

AE : Because these are very complex systems that require multidisciplinary teams. I really enjoy collaborative work. To understand and master all the parameters, you need to bring together different areas of expertise — and that’s what motivates me. The cloud is a playground where everything interlocks: hardware, software, performance, reliability…

PC : What are the main research avenues you are exploring today?

AE : They all revolve around numerical computing. For example, I’m supervising a PhD on heterogeneous architectures based on open-source processors. Today, the market is dominated by major American players like Intel, AMD, and Nvidia. In Europe, we are behind, but open source offers an opportunity to build expertise and develop our own intellectual property (IP) blocks.

We already have solid building blocks, such as the CVA6 processor and the Vortex GPU. What we are trying to do is assemble these pieces to create more ambitious platforms: multi-core, cache coherence, 3D integration, heterogeneous architectures. A roadmap of this kind was recently presented at the ICCAD conference. Another focus area concerns “sparse” data — matrices or graphs where most elements are zero. This type of data is common in the cloud but poorly optimized. We are therefore seeking methods to make computation more efficient and energy-efficient.

PC : You have just taken the lead of the Archi-CESAM project. What motivated you to accept this role?

AE : Denis Dutoit, who initiated the project, did remarkable work. Since my research is closely aligned with Archi-CESAM’s themes, the CEA asked me to support Denis during a transition phase before taking over. I saw this as a real challenge but also a great opportunity. It’s a chance to strengthen existing collaborations, create new ones, and potentially launch new projects in the future.

PC : What are your initial goals as the head of this project?

AE : In the short term, I want to fully understand the workings of the project, meet the partners, and integrate myself. I arrived recently, so I’m still in a learning phase. But quickly, I want to encourage direct exchanges, organize in-person meetings, and engage more closely with everyone. In the medium term, I aim to strengthen our presence at strategic events like the RISC-V Summit. Archi-CESAM is already represented there, but we can go further and assert our leadership.

PC : What opportunities and challenges do you see for Archi-CESAM?

AE : There are already great collaborations underway. For example, a cache memory developed at CEA has been used by the CONVEX team at INRIA for formal verification. That’s exactly what we want: for the building blocks designed here to serve over there, and vice versa. The main challenge remains human: we have planned 12 hires, 8 are already in place. We need to finalize this because a complete team is fundamental to moving forward together. Also, we must stay attentive to the smooth flow of skills and resources between partners.

PC : How do you envision the collaboration between Archi-CESAM members and other projects within the PEPR Cloud?

AE : Cross-project collaboration is a strong expectation from the ANR and also an obvious necessity. We have already started discussions with several partners. I believe this will need to be structured more, perhaps by relying on collective events or shared platforms. This is an area I really want to push forward.

PC : Do you have any specific project management methods or tools you plan to implement?

AE : I really appreciate collaborative tools like the TalkSpirit platform, which makes it easy to share documents and work together. It might sound simple, but it significantly boosts efficiency.

On the development side, I’m a strong advocate for open source. Publishing a component as open source allows you to benefit from an entire ecosystem. For example, the cache memory I mentioned received contributions from people outside the project — it’s a real snowball effect. But we need to find the right balance: not everything can be open. Some elements must remain proprietary for reasons of value creation. Finding that balance is one of the project’s challenges.

PC : How will you evaluate the success of the project in the medium and long term?

AE : Excellent question! The usual indicators are of course important: scientific publications, open-source contributions, deliverables. But I would also like us to be able to measure the quality of collaborations. If the project helps build strong connections and new joint projects, then it’s a success. That’s what will ensure the continuation beyond Archi-CESAM.