A tech & AI journey: 5 years at Enjins with Luuk

Take us back to Day 1: Why did you join Enjins, and what skillset did you have back then?
I did the Master’s in Data Science & Entrepreneurship at JADS, which was practical regarding business context but really missed the engineering practices. We learned to create models and talk to stakeholders, but I lacked the skills to turn that into actual software that keeps running. I had zero experience with concepts such as MLOps, CI/CD, or deployment monitoring. I think things like monitoring might be quite boring for most people… But if you are a bit of a weirdo like me, then you think such things are actually quite cool. I basically wanted to work somewhere where I could learn about this practical side. When Enjins pitched doing Machine Learning from an engineering angle, not just a Data Science angle, this was exactly what I was looking for.

What motivates you to work in Climate Tech?
On one hand, I am just a nerd who loves solving complex puzzles, but for long-term motivation, it helps to know we are doing something cool that has impact. I know it’s cliche… But whether I improve an algorithm by 10% that helps reduce CO2 or build a data pipeline for batteries, I know I am contributing a small brick to solving a huge global problem. You cannot solve the climate crisis alone, but knowing my technical work contributes to something good really helps day-to-day. I wouldn’t want to spend my time only optimizing data science models for advertising revenue. Eventually, you question how useful that really is.

Looking back at your 5 years here, what are the key things you have learned?
It is hard to summarize, but everyone at Enjins is asked to learn very broadly about the IT landscape. You can’t be an expert in everything, but because we do so many different projects, you learn quite a bit about everything. For example, about security, data engineering, and software architecture. I think that is also our strength, we often surprise clients by having strong opinions on CI/CD pipelines or Infrastructure as Code, even though they hired us as “the AI people”. We also have plenty of people here who enjoy diving into a “rabbit hole” about OAuth authentication just to give a demo on it. Sometimes in sales, we encounter topics we haven’t really touched before. But my mindset is usually: we should be able to do this. We just put a few people on it to dive deep, and because we have such a strong baseline, we usually find that we can master that new domain quite quickly.

What has been your most complex project, and what made it such a challenge?
We had to build an image processing pipeline for a factory that had to run on-premise for security and performance reasons. It was a heavy algorithm detecting tiny defects on high-resolution images, requiring parallel processing on their local servers. We had to install and manage a Kubernetes cluster ourselves in a somewhat “old school” IT environment, including hosting the Message Queues like RabbitMQ. We built an auto-scaling system that optimized compute usage based on which step of the factory process was running. It was a cool challenge to see what it takes to build a private cloud situation without having AWS or Azure to fall back on.

How has the Enjins tech stack evolved from the early days to the current landscape?
I think we used to focus on getting a single model running, but now we think more about the entire AI strategy around it. We really try to enable companies to keep working with AI. Also, of course tooling has shifted. From custom Kubernetes setups to better managed solutions from cloud providers, and things like MLflow becoming standard practice. We also see a shift towards streaming technologies and specific tools for LLMs, like LangChain and evaluation frameworks, which didn’t exist when I started.

How do Enjins’ assets like the streaming platform help in your work?
We used to say we built everything custom, which was sometimes an excuse for not investing time in transferability. Now, we recognize that while data differs, practices like “exactly once processing” or monitoring are generic. By formalizing this into assets, like our streaming asset, we give projects a flying start. I think it saves a lot of time on things like setting up monitoring, which is always more work than you expect. It simply allows us to build with higher quality immediately.

How has your role evolved from starting as an AI Engineer to becoming a Lead AI Engineer?
It went really gradually. My role as a Lead is also tailored to my strengths. I knew early on that I wanted to grow technically and didn’t necessarily want to become a manager. As a Lead, I act more as a design architect, focusing on complex technical challenges. I am more often a sparring partner and coach rather than building everything myself. Fortunately, my work remains mostly technical, so I’ve avoided the trap of just sitting in meetings all day. I have learned that you can still channel your technical energy by thinking about high-level design issues without having to type every line of code yourself. It is actually very interesting to think more about high-level design questions, or even philosophically about software architecture as a concept.

As a Lead, what values or skills do you aim to teach the people you coach?
As a lead, you are indeed coaching other engineers. At Enjins, everyone has a coach. I coach the people who often follow a similar technical track as I did. We don’t only talk about technical stuff though, but also personal development (and how you are doing personally). But, I would say the key word I use most often here is curiosity. When we do Design Sprints, people sometimes doubt if they are “expert” enough. I think by simply letting your curiosity run free and asking the right questions, you quickly become the expert in the room. I am also not an efficient machine that walks in a straight line, my brain goes in different directions. I try to teach them that it is actually a strength to be open to diving into a rabbit hole to learn something new, as it might be relevant a year later and helps you grow.

The AI field moves at lightning speed. How do you and Enjins stay on top of the latest developments?
A large part of this lies in the type of people we hire. We select for curiosity across the broad IT and ML landscape. Because we are curious, we aren’t afraid to try new things in projects and share those learnings. We have a culture of sharing, like our Friday afternoon tech demos where we discuss cool tech we encountered. Since our projects are so diverse, we also constantly encounter new domains where we still have plenty to learn.