How and why did the collaboration between ASMPT and Radboud University start?
“Headquartered in Singapore, ASMPT has its roots in The Netherlands. We’re a company in the field of semiconductors, which is quite a vast field. We develop machines that assemble microchips. One of the processes we’re working on is called advanced packaging: we integrate small chiplets onto a 3D pattern so that they communicate with each other to make a functional device.
Since 2017, ASMPT has had an R&D branch located in Beuningen, close to Nijmegen. Our focus there is developing new innovations for ASMPT around the world. To achieve that, we always planned to connect with local knowledge institutions. We originally looked towards technical universities in The Netherlands and other European countries. But since we’re located in the outskirts of Nijmegen, it seemed only logical to see what Radboud University had to offer. We got in touch with two groups: one working on machine learning led by prof. Marcel van Gerven and one on data science led by prof. Tom Heskes. Both were quite eager to collaborate.”
What is the added value for ASMPT to collaborate with Radboud University?
“In our industry, there’s a massive race to shrink elements to micro- and nanoscale. That makes it very complex to identify defects in chips such as cracks or other anomalies. To identify those minute defects in a very complex background, with high speed and high throughput, requires AI. Without AI, you can’t work accurate or fast enough. It’s impossible to inspect everything by people, so you would need to select random samples. By using AI, that is trained on data that is annotated by experts in their domain, and self-develops using machine learning, you reach a much higher level than what is possible as a human being. So we’re not replacing human knowledge but augmenting it with AI.
Together with the researchers at Radboud University, we’re working on two processes. First, identifying defects. Second, controlling the manufacturing process, basically learning what to do with defects and anomalies.
Radboud University has a lot of experience and knowledge in the field of machine learning and brain technology, coming together in neuromorphic computing. The added value of the collaboration lies first in knowledge sharing. We also write project proposals for public funding together. And we work together in an ICAI lab – ICAI stands for Innovation Center for Artificial Intelligence. With 5 PhDs, a number of master students and several professors we work on trustworthy AI models, algorithms and technology for real-time inference, prediction, and control of high-throughput data streams in the semiconductor manufacturing industry. Such a lab is also a place where we can scout for talent, of course. It has led to a very interesting, creative and tight collaboration.
For Radboud University it is interesting to work in this field of semiconductors, which has quite a large presence in the eastern part of The Netherlands. The Dutch government has recently also invested in a strong microchip sector via the Beethoven project. Radboud University can play an important role in the sector, providing talent as well as interesting and innovative ideas that can take us all to promising new directions.”
What kind of entrepreneurs could benefit from collaborating with Radboud University, in your opinion?
“Radboud University is a broad university, so there probably are a lot of fields in which it can play a significant role. From my experience, I must mention the strength of the Radboud University and Radboudumc groups. I can envision a broad range of applications and collaborations, from start up to big industrial parties. The medical arena, precision industries like ours, transportation systems - to name just a few. And of course the clean energy domain, an important societal goal to which Radboud is very committed. Neuromorphic computing plays into that as well: working with very high speed but low energy.”
