sudolabs logo

4. 10. 2023

7 min read

Erik Kandalík: Machine learning beyond neural networks and ChatGPT

With his experience in various fields, Erik has recently taken on a new role, spearheading AI consulting and machine learning services and he is dedicated to educating our team about this topic. How does he do it? And what hurdles lie ahead? He spills all the details in the blog post below.

Silvia Majernikova

Social Media Marketing Manager

Erik Kandalík, seated at the office, with a warm smile, working on a computer and enjoying a cup of coffee

Erik Kandalík, our data scientist, working from Košice headquarters

Let's get right to your passion for AI and machine learning. How did your journey into this field begin?

To begin with, I wouldn't say that my passion starts or ends with AI and machine learning; it's just one of the many areas I find incredibly fascinating. When discussing machine learning, my journey essentially began with a personal interest in the field. I delved into several books like “Deep Learning” (Ian Goodfellow, Yoshua Bengio, Aaron Courville), experimented with various projects like numbers detection with MNIST dataset, or plants classification with Iris dataset, and soon found myself as a member of a data science team, where I gained most of my experience. I consider myself fortunate to have been surrounded by exceptionally talented individuals with diverse expertise in areas such as natural language processing, image processing, signal processing, and more. Now, let's talk about the 'passion' aspect. I've always had a deep-seated passion for problem-solving, and machine learning has provided me with the means to tackle complex challenges. This is where my passion truly thrives.

Erik Kandalík and his loyal Golden Retriever, Gaston, enjoying a peaceful moment in nature

You have a background in working on various projects within a corporate environment. What inspired you to join Sudolabs and how does work here on startup projects differs from corporate one?

First of all, I would like to emphasize that both corporate and startup environments have their respective pros and cons, and the choice between them depends on individual preferences and needs. Personally, my decision to join Sudolabs was primarily motivated by the significant opportunities for career growth that I saw here. I also believe that my rather unconventional background, which spans from web development to machine learning, could be of great value to a smaller company like Sudolabs.

As for the differences between the projects, I would say that the startup world is quicker, and more dynamic, with a strong emphasis on efficiency. On the other hand, based on my experience, the corporate environment tends to move at a somewhat slower pace, prioritizing stability and planning. Corporate settings may sometimes engage in excessive planning, while startups may lean towards a "just do it" mentality. In my opinion, the ideal approach often lies somewhere in between.

At Sudolabs, we are implementing AI consulting/machine learning services. How do you go about fostering knowledge-sharing and collaboration among team members?

My mission, alongside my colleagues, is to educate individuals in the field of machine learning because I believe it's a topic that every developer should have at least a basic understanding of. Machine learning not only empowers us to solve more challenging problems, like recognizing faces or identifying people by voice but also encourages people to approach various issues from different perspectives. Just like any other tool, having a broader set of tools at your disposal enables you to tackle more complex tasks.

To achieve this goal, my colleagues and I started multiple initiatives. I have conducted several internal tech talks covering a range of machine-learning topics, aimed at enhancing overall understanding of the subject. Nowadays, many people tend to associate machine learning solely with neural networks, or even worse just with ChatGPT, but we want to demonstrate that machine learning encompasses a much broader scope. Additionally, we organize workshops aimed at practically utilizing different techniques from machine learning to solve real-life problems. We also recently started moderating a course that specifically focuses on deep learning.

Erik Kandalík and Tomáš Kuchárik, data scientists, collaborating at work

You mentioned workshops. What are they like? Do you have any examples of resources you use?

As I mentioned earlier, the primary goal of our workshops is to bridge the gap between theoretical subjects, particularly in machine learning, computer science, and real-life problem-solving. Ideally, we aim to address problems that are relevant to our company's internal operations or have connections to our customers' needs. While learning about theoretical concepts can be valuable, we've found that practical experience is more engaging for a broader audience.

Now, what are these workshops like? Time consuming. Designing a workshop for individuals with varying levels of expertise and diverse backgrounds, while keeping it engaging and connected to real-life problems, is a considerable challenge. Often, compromises have to be made to ensure delivery promptly.

When it comes to resources, the selection depends largely on the workshop's topic and the attendees' expertise. For instance, we held a workshop focused on clustering algorithms, where the algorithms are well-studied, so it was quite easy to find relevant resources in the books or online. There are so many resources that we had to narrow the scope to focus on the main ideas, but we encourage participants who wish to explore the topic in greater detail to pursue further knowledge individually.

If there is someone who reads this blog and would like to get into the data science field, I would suggest a few good resources in my opinion, For anybody who appreciates a practical approach the course from Jeremy Howard is for you. In case someone is looking for more theoretical courses, there is a lot to choose from. For the NLP I would suggest “Natural Language Processing with Deep Learning” from Stanford University and for the more general theme I think “Introduction to Deep Learning” from MIT is of very high quality in my opinion. (even though I didn’t watch the full course, so take it with a grain of salt).

What challenges do you see in educating others on machine learning?

Well, everyone is unique, and their motivations for learning machine learning vary greatly. Some people are primarily interested in the practical applications, while others prefer to delve deeper into the theoretical aspects. Some seek immediate results, while others are driven by curiosity and a desire to understand the inner workings.

The challenge lies in accommodating these diverse preferences and motivations. It's nearly impossible to satisfy everyone's specific learning style and goals. However, the good news is that when people have a genuine desire to learn, these challenges become more easily manageable.

Erik Kandalík and Tomáš Kuchárik, data scientists, engaged in a collaborative work session in our kitchen's common area

With AI booming, a lot of companies are trying to apply it to their businesses. What would you advise those only starting with it?

In light of the recent popularity of large language models and the growing interest in AI in general, it's important to address certain misconceptions that can arise. Some might believe that these systems can effortlessly solve every problem, replace every job, and automate every aspect of their business operations. However, if your company is considering AI adoption for whatever reasons, I would advise approaching this with a degree of caution, especially in the current landscape.

How does your work benefit clients?

I believe that sharing knowledge across the whole company opens new ways of thinking, enabling everybody no matter what project to spot the opportunities when they are present, simply knowing about the options. This might be as easy as recognizing that we can improve search on the website with semantic similarity or making data analysis on customer data for him/her to gain a competitive advantage.

The thing is, machine learning enables us to solve more complex problems, problems we might have thought were unimaginable before. I do my best to help our clients solve these problems, with the hope of consequently improving their business, either directly or indirectly through knowledge sharing in the company, whether it is done with workshops, tech talks, or during coffee breaks. 🙂

What is your guilty pleasure?

Well, when it comes to guilty pleasures, two things immediately came to my mind. The first one's pretty common, I guess – I'm a foodie. I enjoy good food to the point where my restaurant bills get quite large. And once your restaurant bills exceed the mortgage, it's time to do something about that. So nowadays, I'm making an effort to stay within a reasonable budget.

Now, the second thing is a bit more unique, I suppose. I genuinely enjoy learning new things. To the extent that my girlfriend occasionally pokes fun at me, saying I take vacations just to go back to school. Yep, I'm one of those perpetual learners.


Let's start a partnership together.

Let's talk

Our basecamp

700 N San Vicente Blvd, Los Angeles, CA 90069

Follow us

© 2023 Sudolabs

Privacy policy
Footer Logo

We use cookies to optimize your website experience. Do you consent to these cookies and processing of personal data ?