Mads Villadsen is the UNSILO CTO. This means he has the responsibility for selecting which technologies are used, overseeing the overall product roadmap, and, most importantly, managing a team of highly creative experts in machine learning and related technologies.
What were you doing before joining UNSILO?
I worked at the Aarhus University Library for twelve and a half years before joining UNSILO. Aarhus is an unusual library in that it is both a university library and one of two Danish national libraries. As a national library, we collected not only books, but also newspapers, recorded music, and radio and TV programmes. It has some impressive statistics – around 18 million newspapers with full text searchable, 1.2 million TV programmes, 700,000 radio programmes, plus some TV commercials for good measure.
What do you think was your major achievement at the library?
We built a library search system, some years before the launch of the major commercial discovery systems now available (such as Summon, EBSCO Discovery, etc). This system had the challenge of indexing many disparate types of information, using different formats (printed books and journals, as well as online content) and very different metadata standards – what is often solved by using a federated search system (searching each of the repositories in turn). In all, we identified ten different kinds of resource. We replaced federated search by what we called “integrated search”, which was our creation: instead of repeating one query across many systems, we aggregated all the data in a single repository, and so provided proper relevance ranking from one search query. We called our system “Summa”, and we were very flattered that ProQuest, who came to view our system, later called their commercial discovery service “Summon”. The current university search system, which includes elements from our original Summa vision, can be viewed here. Today, the front end is created by the library, but the back end combines elements of the two systems, Summa and Summon.
Building an entire system must have been quite a challenge! What did you learn from this initiative?
I learned several lessons from this project: how to deal with developers, of course, but also how to have a vision and actually execute it; how to work directly with end users; and not least, how to say no to features that aren’t relevant.
How did you find out what the users wanted?
We carried out lots of studies, including interviews and surveys, mainly of undergraduates. One problem of working in a library is that the librarians are not the users. They interact with the users, but they are operating at a different level, and they cannot easily communicate with their users directly.
How does working at the Library compare with UNSILO?
Well, a University Library is an older institution; this means the processes for decision making are slower and things are done by precedent. There was a lot of discussion about how to facilitate technical innovation. That’s not an issue at UNSILO! Here at UNSILO, there is a great vision and we are certainly ahead of the curve (while at the library we were catching up in many areas). The people are equally dedicated in both places, but the infrastructure, the way of doing things, is very different. Incidentally, when student look for documents in a library catalogue they are have the impression that the cataloguing information is always correct, which of course may not be the case.
The team you manage here comprises world-class experts in machine-learning and natural language processing. Don’t you find it a challenge keeping them all working to a common goal?
The beauty of being a startup is that everyone is dedicated to the vision, not to work 9-5 in their own domain. Everyone works to identify concepts across several subject areas.
The University Library comprised several hundred staff, while at UNSILO you have a much smaller team. What size organizations do you prefer working with?
Actually the development team I worked with at the library was a similar scale. The difference is that at UNSILO there is less of an internal structure to navigate – especially as the library was the primary customer for the product we were developing – we were developing a product for colleagues who you will be having lunch with the next day!
At the University, you were providing services for local and national users, but here at UNSILO, the situation is the opposite – pretty much all your customers are international. Isn’t that a challenge?
In the modern world, it doesn’t matter – it helps simply that there is an overlap in time zone which means we can communicate easily. There is some travel but it is mostly within Europe. There is an additional level of coordination required with people working remotely, but many of our customers are fortunately very understanding about time zone difficulties in arranging meetings.
Many thanks for your time.