How creating a Science Venture allows to scale impact without selling out our values
ASReview LAB is free, open, and fully funded. Yet, to truly serve the community, we needed a vehicle that could move forward faster than a university department ever could. Here is my side of the story about why starting a Science Venture helped to bridge the gap between academic innovation and real-world implementation.
In my previous post, I shared the story of how ASReview LAB grew from a bored researcher’s daydream into a global open-source ecosystem, fully funded by the VICI and other NWO grants. But success creates its own gravity. As the AI-lab at the university grew, we started receiving many requests that went far beyond the core values and capabilities of a university. We realized that if we wanted to truly impact the world, we needed to solve three specific problems that a traditional academic lab simply isn’t built to handle.
Barrier #1: The Support Gap
Large organizations and industry players are eager to use ASReview, and they absolutely can! After all, the software is open-source and free to download.
Technically, running the server stack isn’t the issue; internal IT departments or third-party IT providers can usually handle the hosting. The real bottleneck is the implementation strategy.
A university lab is an engine for innovation, not implementation. We are not an IT helpdesk, nor are we a software vendor. Implementing ASReview within existing corporate pipelines presents its own set of complex challenges: Which model is best suited to this specific proprietary data? How do we connect the API to other internal tools, and how do we validate the workflow for compliance?
When companies ask these questions, they are asking for professional services that a PhD student or a professor cannot (and should not) provide. Therefore, we needed a team dedicated to the practical side of running studies within organisations, specifically so the researchers in the AI-Lab could stay focused on what they do best: scientific discovery.
Barrier #2: The Agility Crisis
Simultaneously, we faced a fundamental mismatch in timelines. We received constant requests for short-term, high-impact projects, validating AI pipelines, testing models on private internal datasets, or running quick screening sprints. For these partners, the goal wasn’t a scientific paper; it was immediate, actionable results.
The problem is what I call “University Time.” With current budget cuts in Dutch universities, hiring permanent staff for these specific roles is impossible. Even when project funding is available, recruiting skilled temporary staff with high-level AI expertise is incredibly difficult and slow. My only other option was “buying out” teaching time, but the rigid educational timeline (semesters) clashes with the fluid innovation timeline (weeks). Obviously, I cannot cancel a course for my students next week just because an urgent project lands on my desk today. Also, a client does not want to wait till next year before we can arrange some time from a university teacher.
The consequence was painful: we were forced to say “no” to exciting, impactful work, not because we lacked the knowledge, but simply because we couldn’t mobilize people fast enough.
Barrier #3: The EU Bottleneck
Finally, our success led to invitations to join numerous EU consortia. However, the requirements often shifted from pure research to production-level infrastructure, such as building knowledge warehouses for specific grant topics. This is essential work, but it is production work that lies far beyond the scope of a university. Simultaneously, we were asked to collaborate on emerging AI frontiers, validate Large Language Models (LLMs) for tasks related to our work in the AI-lab but requiring different infrastructure, and implement AI at scale across many different consortium partners.
While prestigious, these grants require extreme administrative agility and fast turnaround times, often falling during the holidays. Our university’s Research Support Office (RSO) is excellent, but its capacity is naturally finite. They simply couldn’t process the sheer volume of participation requests we received. (To my dear RSO support staff: if you are reading this, thank you for everything you have done already! I know I overloaded you!)
We risked becoming a bottleneck in these international collaborations, potentially missing out on shaping these new directions in European AI innovations.
The Solution: A Science Venture
To break through these walls, we sat down with the Rector of the university and the Dean of our faculty to design a structural solution. The result was the creation of a Science Venture FwdFaster AI, together with Jonathan de Bruin and Frans Folkvord, independent of the university.
It is important to make a distinction here. We did not create a traditional “Spin-off” to “sell out” or exit the university. Instead, we built a Science Venture to “scale out.”
- Support: A dedicated commercial team now handles the testing, implementation, and support for complex in-house validation workflows.
- Agility: The Venture operates like a startup. It can hire talent quickly, deploy staff to short-term projects instantly, and doesn’t need to wait for a new semester to start working.
- Capacity: It acts as a professional partner in EU consortia, handling its own administrative load without burdening the university’s Research Support Office.
A Symbiotic Ecosystem
The result is a symbiotic ecosystem in which I work part-time for both entities. At the Science Venture, we handle implementation, speed, and future AI developments. At the University’s AI-Lab we handle the fundamental AI development for ASReview and its maintenance.
This distinction is vital: ASReview remains at the university. It is secured by NWO grants and will remain open, free, and accessible to everyone. By offloading commercial pressure and administrative complexity to the Venture, the academic team at the AI-Lab is free to continue doing what we do best: teaching the next generation of students and researching the fundamentals of AI.
A crucial clarification: If you are simply looking for a software company to host ASReview on a server with 24/7 IT support, please look for an IT provider; there are plenty of excellent ones out there. That is not what we do at FwdFaster AI!
But if the Venture doesn’t offer simple hosting and doesn’t own ASReview, what exactly does FwdFaster do? I will explain that in the next blog post!