Employee Spotlight: Evertjan Peer

Meet Evertjan Peer, Tech Lead at KickstartAI. From operations research to ecommerce and startups, Evertjan brings deep technical expertise and an entrepreneurial spirit to our team.

Let's dive into his journey into AI, what drives his work, and the lessons he’s learned along the way. 

Hi Evertjan, what's your role at KickstartAI?

I'm the Tech Lead. This means that I’m responsible for the technical execution of our partners' projects — from organizing the challenges to the implementation phase of the AI solutions we develop. I also lead the work of our engineering team. 

Can you tell us about your journey into AI?

My path started at the intersection of operations research and computer science — I pursued both master's programs simultaneously at the Eindhoven University of Technology. For my final graduate project, I worked with Dutch Railways (NS) on applying deep reinforcement learning to train shunting problems. Essentially, this was about planning how trains move around shunting yards in the evening so they can leave in the right order the next day.

After that, I wanted to see AI applied in industry. I also wanted to work abroad, so I joined Zalando in Berlin. I spent five years there, first working on traditional data science projects like customer segmentation. I then moved to the recommendations team, where we personalized the entire shopping experience based on user behavior.

Following my years in Zalando, I co-founded a startup that focused on voice-based administration, with an initial focus on healthcare. We built an app to convert spoken words into database entries for healthcare professionals. While we had some initial customers, it didn't take off like I had hoped. But it was a valuable experience. And it led me back to the Netherlands and to KickstartAI, where I started as a senior ML engineer before becoming tech lead.

What project are you most proud of?

I'm overall proud of our approach to the challenges and projects we do. The NS challenge and Ampère challenge, for example, show how the different parts of KickstartAI work together — our founding partner companies, our community of engineers, and real technical implementation.

These aren't just hackathons where we let people loose. There's careful preparation to ensure we reach a technical implementation which is actually used.

During these intensive three weeks, there is just so much energy in the teams. Many participants tell us it feels like a startup, which is refreshing for those coming from larger companies.

With the Ampère project, we went from challenge to implementation and real-world impact in two and a half months. That was really cool to see!

What drives you in your work?

Two things. First, I need to know how my work brings real value — it's very hard for me to do experiments for the sake of experiments. I always think about how we're helping people. That's why I love that all our projects are tied to solving actual problems.

Second, there's this fascinating element of mystery in machine learning that really grabs my attention. In operations research, the focus is on statistics and optimization. You find the best solution and know exactly why it's optimal. But with machine learning, especially deep learning, you don't always know how the decision is being made. That “mystery” aspect, particularly with recent GenAI advancements, makes it very attractive for me to spend all my time in this field.

What AI tools do you use most in your daily work?

I always pay for one of the major chat models — either Claude or GPT — and alternate between them to see how they're progressing. They're my general assistant. I use them for everything, from reviewing house-hunting paperwork to helping with unfamiliar tasks in marketing or legal.

I also use Aider, a terminal-level AI programmer, to build tools and test the boundaries of what's possible with AI coding assistance.

Recently, I went to a silent disco event. The next day, I wanted to replay the playlists from the different channels that were powered by the attendees’ own Spotify accounts. Using Aider, I built an app in an hour and a half that could create Spotify playlists based on playback history from specific time ranges. It cost me €2 in tokens, and it was amazing to see — because this kind of development used to take much longer. It's a great way for me to see where we stand with AI capabilities.

What do you enjoy most about working at KickstartAI?

I really enjoy being partners, not just external consultants, with the companies we work with. I love that we're at the forefront of new developments — there's no day at KickstartAI where we don't talk about the latest models and advancements.

Also, being in the centre of this AI community that we've built ourselves is special. We're always surrounded by inspiring people, whether at our meetups or working with teams at our partner companies. People who are really at the forefront of their field.

What advice would you give to aspiring AI professionals?

It's crucial to understand the problem you're trying to solve. Talk directly with the problem owners. Understand the nitty-gritty details of how people will use your solution. This helps with defining metrics and evaluating your solution. It also gives you vital background for assessing your data quality. The further you are from actual usage, the easier it is to make wrong or costly assumptions.

Fun fact about yourself?

I sometimes wear toe socks! After reading "Born to Run" and getting interested in barefoot running, I started wearing individual-toe socks under my regular shoes. I’m not brave enough for those five-finger shoes yet. So, I stick to toe socks as my little secret!

💡 Want to learn more about KickstartAI and our mission to accelerate AI adoption in the Netherlands? Check out our website or connect with Evertjan on LinkedIn!