Close Close icon
Close Close icon

Learn & Work

Whether you are learning, studying or working, Brainport Eindhoven offers limitless opportunities for growth. Your success depends on the way you overcome your challenges. Please contact Brainport if you need any support. We will help you gain more knowledge and new perspectives or we will just answer your questions.

Close Close icon

Latest news

12 October 2020

DSM and Lightyear aim to design solar roofs for all types of electric vehicles

  • Mobility
Arrow icon All news

Upcoming event

Arrow icon All events
Close Close
Close Close

Dream big but start manageable

Technology Owner at Vanderlande explores and demonstrates the power of AI.


Mariana Goldak-Altgassen from Ukraine works at Vanderlande as Technology Owner Algorithms within the Innovate department. In this position she is in the driving seat to demonstrate both the power of AI and its limitations to the internal organization and the company’s customers. In the short-term Mariana expects a lot from advances in computer vision area. The first proof of concepts (PoCs) are very promising. The main challenge lies in scaling-up. Vanderlande actively collaborates with other companies and universities to build up knowledge and speed up the industrialization process. Mariana’s main advice when it comes to exploring AI and defining use-cases: ‘dream big but start manageable’.

“I came to the Netherlands in 2010 for a Professional Doctorate in Engineering program at TU/e called Mathematics for Industry,” says Mariana. “After I had finished the program in 2012, I joined Vanderlande, a worldwide leading supplier for sustainable logistics process automation in the airport, warehousing, and parcel market. At Vanderlande I started working at the R&D department as a software development engineer.”

Fascinated by machine learning and AI

“While I was working on a project called Load Forming Logic, which formed the heart of our STOREPICK solution, I became interested in Machine Learning. At that time, we were exploring options to predict stack-ability of items. I found it really fascinating and that is where my journey began. I started taking various courses in the field of Machine Learning and applying these new skills. At a certain moment in time my role changed to a somewhat more managerial position. I was offered a position as team lead of my team.”

Responsible for AI roadmap creation and connecting to the ecosystem

“I took on the challenge and as a team we started to build up AI competences. As the Vanderlande organization changed in structure, my role also changed into a more content-oriented role called Technology Owner. In this role, I am responsible for defining the AI technology roadmap, coordinating activities with academia, supporting the development organization with the transition of technology when it becomes more mature and ensuring new technology is anchored in the entire organization.”

Exploring the advantages of AI through use-cases

“Our Innovate Department is in the driving seat to show both the power of new technologies as well as their limitations. A missionary role to show the company what each technology means. At Innovate we therefore focus on the front-end of things, the proof of concepts. When a proof of concept is a success, it will be productized.”

Successful PoCs Focus lies on applying computer vision

“We did various exploratory projects to see what the value would be of applying AI on certain applications. For example, instead of using classical vision algorithms, and expensive hardware, we were able to use smarter algorithms and less expensive hardware, without sacrificing the performance of parcel detection and thus achieving significant cost-reduction.

Baggage singulation, item picking from a bin, parcels detection on a conveyor, operator’s posture analysis for ergonomics feedback are all examples of computer tasks for which AI allows us to offer more versatile and robust solutions to our customers. This is especially important in the light of the ever-growing demand and diversity.”

Scaling-up successful PoCs

“In general, we explore a lot. We have successful proof of concepts, the following step is to take them to the next level. Vanderlande and other companies are still learning how to industrialize AI solutions. It concerns everything related not to AI itself but the whole infrastructure around it, from gathering data to deploying models, maintaining them over time, detecting the drift of a model and acting upon that. It is something that still needs to be addressed. That is the phase we are now at.”

Aim is demonstrating the value of technology

“At Vanderlande we are still exploring the value that AI can bring, our main aim is to demonstrate that value. You can define a hundred use-cases, we are focusing on the most promising ones. In our department we work according to a clear process to make sure that what we work on is aligned with what the business needs. The technology owners look at the latest trends, evaluate what is technically possible and work closely together with people who have a good understanding of our customers’ needs. All ideas are carefully assessed, and the most promising ones continue to the next phase.”

Broadening the scope of AI

“Besides advanced computer vision we are also exploring other areas. If you compare AI to a human, you have eyes, hands, brains etc. On the short term we will use AI for “eyes” as there are so many use-cases in the vision area. AI allows our systems to understand what the cameras see and thus provide perception capability that enables our systems to deal with more diverse and less structured environments, which was much harder or even impossible to achieve with classical approaches.  In the long run, the biggest impact will be in the “brains”: on a higher level of our systems, to control all the processes that our systems are responsible for by decision making algorithms.”

Self-aware systems that make autonomous decisions

“The holy grail solution would be self-aware systems, systems that have all the sensors needed to gather the right data and then you use these with AI on top of it to make the best decisions. One trend we are working towards is dynamic decision making that we can, for example, use for order batching, vehicle routing, stacking and other optimization problems. You can preplan some things, but the world is dynamic and in reality, we need to act upon actual situation of the environment. To enable real-time adjustments to our systems with minimum impact on the system down-time, it would be very beneficial to have an accurate digital representation of such a system (with a connection to the physical system). Then, we can optimize the parameters, validate impact of the changes and only then apply them to the physical system with high confidence.”

AI can empower people

“I believe that AI is here to empower people. You need a human in the center and use AI to make the human shine. It is not a goal to replace humans but make their lives easier. Repetitive, boring tasks need to be automated but interpreting, showing emotions and being creative, remain unique human capabilities.

There are already some AI applications to automate activities for engineers on site. For example, if you have a chatbot on site, the bot can guide your maintenance and, in this way, make it more efficient, you don’t have to read the entire manual then.”

Collaborating with EAISI on deep reinforcement learning

“To bring AI further we collaborate with various universities. We are in the process of starting a consortium with several industrial partners and TU/e, EAISI. In this consortium we will explore deep reinforcement learning, one of the techniques that fall under the AI umbrella. The objective will be to develop a generic framework for optimizing decisions in dynamic environments in logistics, which is supposed to make it easier to implement these kinds of algorithms in industrial settings.”

Consortium aimed at developing a trustworthy and responsible AI framework

“Another consortium that we are preparing for will start in 2022, if it will be granted by our national research council NWO. Jheronimus Academy of Data Science is in the lead, even more industrial partners are involved in this. The objective is to develop a trustworthy and responsible AI framework to augment human operator skillsets and achieve hybrid intelligence in the end.

Besides participating in consortia, we work together with very diverse companies that we exchange knowledge with, or that provide solutions to our needs.”

Challenges in further developing AI

“In further developing AI we face different challenges. As I already mentioned the industrialization of AI solutions is an important one. Next to it, making sure that we have the right data in of good quality is also challenging. Sometimes you already have data that you can use, but it might turn out that what has been collected does not provide sufficient input. Then you need to think about what data do I need, and accordingly you need to collect it. Important is to do it in a responsible way, since collecting all possible data is not efficient and it also has negative carbon footprint on our planet.”

One of AI’s main limitations is that some techniques, like deep learning, is a black box, when you compare it to classical software. If you need to drill down the exact root cause, it might not be the right technique. On the other hand, there are also simpler Machine Learning (sub-field of AI) techniques, like decision trees, which are explainable and could be used to fulfil the task. There is a lot of work that needs to be done at an academic level to increase explainability of AI.”

Creating the right amount of understanding of AI at the right level

“Among other non-technical challenges, lies understanding. There is a lot of explanation to do for us internally, to explain what AI is and what it is not, when it does bring value and when it is not the right tool. It would be good for any company in our region to have a global training program, to ensure that employees improve their understanding of the technology according to their functions/roles.”

Dream big but start manageable

“Finding experienced AI people is not easy, they are quite scarce. We should not expect to find gurus who have years of experience. Instead, provide training to our current engineers and focus on education programs.

Investing in AI knowledge and skills of students coming from universities and schools is important. We should address it at a fundamental level, making sure that young professionals learn relevant things, to be fit for the future.

Finally, my advice is to dream big but start manageable, by focusing on most impactful use cases. We don’t have an army of AI engineers to work on everything that is possible. Possibilities are limitless.”