DATA STORIES | GAMIFICATION | KNIME ANALYTICS PLATFORM
Learn From The Best: The Finalist Teams of KNIME Game of Nodes
My Data Guest — An Interview with Team AST and Team FI-Reunited
It was our pleasure to interview the finalist teams of the KNIME Game of Nodes tournament as part of the My Data Guest interview series.
KNIME Game of Nodes is an international tournament where teams of 2–3 people from the KNIME community compete against each other to solve data analytics challenges and showcase their data science skills using KNIME Analytics Platform. This year, the challenges ranged from training predictive models and web scraping to engineering AI-driven solutions, plotting geospatial maps, designing recommendation systems and much more.
Ultimately, at the finale played live at KNIME Spring Summit 2024 in Austin (TX), Team AST emerged victorious, earning the prestigious Golden Node.
We discussed with the finalist teams their professional careers, we talked about their encounter with KNIME, and asked them to share their strategy as well as tips and tricks on advancing successfully in the tournament.
Our Guests
Representing Team AST:
- Takanobu Araki, Senior scientist and Manager at a Japanese Pharmaceutical company
- Toshiyuki Ohfusa, Data scientist and Manager at a Japanese Pharmaceutical company
Representing Team FI-Reunited:
- Martin Dieste, Head of Finance at Orbit Health
- Reza Tavakol, Credit & Equity Risk Controller at Siemens
We’d like to send our warmest congratulations on their journey to the team members of Team AST and Team FI-Reunited who could not be with us on this My Data Guest episode.
Roberto: Can you tell us more about your professional selves?
Martin: I’m the Head of Finance at Orbit Health, a digital health startup. My day-to-day responsibilities cover everything from managing bookkeeping to overseeing controlling and reporting tasks. KNIME has been an invaluable tool in our setup, especially for generating reports efficiently.
Reza: My background is in electrical engineering and computer science and now I’m in risk management at Siemens Bank, where I dive into data to manage risks and support system migration.
Takanobu: As a researcher in a Japanese pharmaceutical company, I use KNIME to analyze in-house data, aiding compound design. My goal is to create tools that can accelerate drug discovery.
Toshiyuki: My background is also in medicinal chemistry but I transitioned to being a data scientist in a pharmaceutical company, where I also focus on drug discovery.
Roberto: How did you get into data analytics?
Martin: My background is in business and management, and back then, digitalization wasn’t as prominent as it is today. I’ve always had an intrinsic interest in data. I started with automation using Excel and VBA, and then later on I moved towards Python, KNIME, and Power BI. You can say I was an early adopter.
Reza: My interest in data science emerged from a combination of my academic background in computer science and my work experiences, with a particular interest for business intelligence. This blend introduced me to tools like KNIME and Python.
Toshiyuki: I didn’t have any formal education in data science. However, during my time as a medicinal chemist, I experimented with AI and data in research automation. This experience led me to utilize data for synthesizing new chemical compounds and accelerating drug discovery, making data analytics a significant part of my current work.
Takanobu: I have gained my skills in data analysis through work rather than formal education. I often explore data to consider what kind of compounds to design and synthesize next. In the past, my data analysis skills were limited, but thanks to KNIME I can now conduct more extensive data analysis.
Roberto: Can you share a critical use case where data-driven approaches are crucial in your field?
Martin: Data plays a crucial role in digital health start-ups, informing financial decisions and business case calculations. One solution I’m particularly proud of involved extracting historical data on disease diagnosis and hospital admissions from thousands of hospital quality reports in XML format to aid my colleagues make informed decisions.
Toshiyuki: In drug discovery research, we heavily rely on data to design new molecules. The use of machine learning and AI is a hot topic, helping us create better molecules faster and bring them to the patient bedside efficiently.
Roberto: How did you encounter KNIME?
Martin: I stumbled upon KNIME in 2018 during my time at Siemens Healthineers. It wasn’t until Reza joined and showcased some impressive work that I decided to embrace the low-code approach.
Reza: I encountered KNIME in 2020 thanks to the collaboration with great colleagues, such as Martin, who I want to thank publicly for his mentorship. I explored the new world of possibilities that a low-code/no-code tool offered for automation and analytics.
Takanobu: I discovered KNIME two years ago, as my role transitioned to one focused on data. KNIME was already used internally at my company, providing a perfect starting point for learning and integrating programming and data analysis.
Toshiyuki: I came across KNIME four years ago, as it was introduced to me by a colleague researcher. After that, I kept learning through a Japanese blog and self-studied the tool functionalities, leveraging community forums for support.
Roberto: As finalists of the KNIME Game of Nodes, you’ve shown exceptional data science skills and KNIME proficiency. How did you become so proficient in the software?
Martin: I started with a basic self-paced course to understand the interface, then learned on the job by solving real problems. I used YouTube, the KNIME Forum, and examples on the KNIME Community Hub. Very recently, I took the opportunity of the 90-day-certification challenge to put my knowledge to the test and obtained up to L4 certification in some specialities.my That was really a great initiative from KNIME to encourage users to get certified for free!
Reza: I can relate to what Martin said. Apart from a few YouTube tutorials, I mainly learned by getting hands-on experience. Searching the KNIME Forum and Google for specific use cases helped me find and learn from others’ workflows. Getting my hands dirty was key for me.
Toshiyuki: I learned on my own using KNIME’s resources and Japanese blogs. Most of my questions were promptly answered in the KNIME Community Forum. Constructing workflows, even if at first it can be a bit time-consuming, is the key to proficiency.
Takanobu: I learned the basics from KNIME resources in Japanese and from our team leader. After that, I improved my skills through practical trial-and-error at work, and by participating in the Just KNIME It! challenges and other gamification initiatives like this year’s Game of Nodes tournament.
Roberto: What are your favorite KNIME features or nodes?
Martin: I love the Rule-based Row Filter and the Column Expressions nodes for their flexibility and versatility. Very recently, I got to love the Generic ECharts View node and its integrated AI assistant to effortlessly generate JavaScript code and customize plots.
Reza: My favorite KNIME features echoed Martin’s favorite. In addition to that, I’d like to include components for their re-usability and ability to scale up projects easily, the REST Client extension to interact swiftly with APIs, and the Python Integration for more experimental and highly technical projects.
Toshiyuki: Components are also on top of my list, for it’s so easy to create an interactive dashboard with them. Additionally, as a drug discovery researcher, I really love the RDKit-related nodes to analyze chemical structure data. The last one is the Bash node. It is very helpful to connect some outer specialized software or scripts.
Takanobu: As for me, I really like components as well, the nodes where I can enter RegEx such as the String Manipulation, and the possibility of embedding the data in the “workflow data area” to ensure portability.
Roberto: Let’s now talk about your stellar journey in the KNIME Game of Nodes tournament. What aspect of the competition did you enjoy the most?
Reza: We had a blast throughout the competition, as it was also the occasion to bring together ex-colleagues, who don’t work together anymore. That’s indeed why there’s the word “reunited” in our team name. On top of that, we really enjoyed the learning opportunities that the challenges provided. We got to experiment with features of the software and ventured into topics we had not used KNIME for, specifically machine learning and sentiment analysis. Lastly, we really enjoyed the collaborative effort to deliver high-performing and innovative solutions that stood out from the rest, especially because we were battling with people we didn’t know but were equally passionate and eager to make it to the next round.
Toshiyuki: We really enjoyed the tournament and had a great time solving the tasks. Also for us, the learning opportunity that each challenge offered was the aspect that we appreciated the most. Especially because we could apply techniques and work with data types that are different from what we usually work with in drug discovery research. Additionally, we also had fun playing with GenAI to create our team image, trying out different styles and prompts.
Roberto: Which challenge did you like the most and why?
Reza: I enjoyed the sentiment analysis challenge the most. It was my first time using KNIME for this, and I found it particularly intriguing. Despite some technical difficulties, like my laptop almost overheating while tweaking hyperparameters, it was a rewarding experience that tested my technical skills and resilience.
Toshiyuki: I really liked the challenge about creating geospatial maps of US cellular tower coverage and defining an investment strategy for the telco network. It was fast-paced and we were impressed with KNIME’s capabilities.
Roberto: And the challenge you liked the least?
Martin: All challenges were well-designed but if I had to pick one I’d go for the round of 16 where we had to build a stock ticker dashboard. As someone who works in finance, it was less exciting for me since it felt like my daily work.
Toshiyuki: Predicting the annual salary of a data scientist was tough. We found a bug in the workflow, our team leader and Takanobu had to fix it at midnight while I was trying to help but I was at a laundromat because my washing machine had broken on that very same day. The typical situation where Murphy’s law would apply!
Roberto: What was your strategy for tackling the challenges so effectively? Did you appoint a team lead and assign tasks?
Martin: It depended on the challenge. We identified what we needed to learn and what we already knew. We aimed to solve the challenge and differentiate our solution with unique features. We’d start with a Monday meeting, experiment individually, then reconvene to finalize our approach and ensure we had a polished submission by the weekend.
Takanobu: Our team leader played a crucial role in designing the workflows. We would work on them independently, and then after a couple of days we would meet to share ideas and assign tasks to refine our solution.
Roberto: Your solutions were always advanced and well-documented. How much time did you invest in building them?
Martin: It varied by challenge. I spent about half an hour to an hour daily, primarily in the mornings before work, and more time on weekends, especially Sunday evenings for polishing and documenting.
Toshiyuki: We spent 3 to 5 hours on weekdays, and 7 to 12 hours on weekends. Our team leader, in particular, is a great KNIME fan and was always working on KNIME.
Roberto: What is the most valuable lesson you learned from this experience?
Reza: Going the extra mile is worthwhile. Teamwork, communication and adaptability are crucial too, if you want to enjoy a sustainable process and obtain good results. Additionally, before starting a project, it’s a good idea to see if someone else has created a similar workflow on the KNIME Community Hub or on the KNIME Forum — it can save a lot of time.
Takanobu: Teamwork and good leadership are great engines to move forward successfully. Additionally, it’s important to stay focused and remember the original goal and purpose of the workflow, especially when tasks are exciting and there’s the risk of getting carried away.
Roberto: We’re planning the next edition of Game of Nodes. What advice would you give to future participants and us as organizers?
Martin: For participants, just go for it — you have nothing to lose and much to gain. For organizers, thanks for a great experience. It would be helpful to see the challenges and scores of other teams after submissions to understand what sets apart a winning solution.
By the way, after the tournament ended, we shared our solutions publicly on the KNIME Community Hub for everyone to enjoy, learn and explore.
Toshiyuki: I’m really thankful to the organizers who were able to orchestrate such a wonderful tournament, so I would really encourage anyone to participate. I also believe that sharing the workflows of all teams would also be beneficial for understanding different approaches.
Roberto: Before we say goodbye, how can people get in touch with you?
Martin: The easiest way is to reach out to me on LinkedIn or on the Medium blog.
Reza: Same for me, LinkedIn is probably the best option.
Toshiyuki: You can reach out to me on LinkedIn — happy to get in touch!
Takanobu: LinkedIn is the easiest way to reach out to me.
Watch the original interview with the finalist teams of the KNIME Game of Nodes tournament
