Skip to ContentOctai Logo

Journeying from Winning Kaggle Competitions to Solving Real-World Problems with AI

In Herman Hesse’s The Glass Bead Game, the main hero seeks to become a Magister Ludi, or Master of the Game. A Master must be on the way toward perfection by learning a variety of ... - Machine learning predictive analytics with no-code
13 March, 2023
Est. Reading: 4 minutes

Come Meet the Kaggle Champions of Aigoritma

In Herman Hesse’s The Glass Bead Game, the main hero seeks to become a Magister Ludi, or Master of the Game. A Master must be on the way toward perfection by learning a variety of skills—striving to reach the center—not the periphery. 

It’s the same with Kaggle competitors. Kaggle empowers users to work toward becoming data science masters by finding and publishing datasets, exploring and building models, working with fellow data scientists and machine learning engineers, and entering competitions to solve data science challenges.

Aigoritma is home to one of the world’s 400+ Kaggle Grandmasters and two other remarkable Kaggle Champions. Let’s meet our heroes, learn about their respective  journeys toward the center of data science—and see how Kaggle has helped build their character and skills.

What Exactly Is Kaggle?

  • Kaggle was founded in 2010 and acquired by Google in 2017.
  • Kaggle allows users to collaborate with other users, find and publish datasets, use GPU integrated notebooks, and compete with other data scientists to solve data science challenges. 
  • The aim is to help professionals and learners reach their goals in their data science journey with the powerful tools and resources it provides.
  • Kaggle competitions are one of the sub-platforms that have made it such a popular resource.
  • Kaggle also provides Learn Guides—collections of high-quality learning resources authored by the Kaggle community.

Getting Hooked on Data Science

Güneş Evitan started programming as a hobby and soon discovered data science. The thrill of solving real-world problems with data was incredible—soon he was hooked. Acquiring new domain knowledge to understand real problems is what excited him most.  

Güneş started taking part in Kaggle competitions to learn from the community and for the chance to compete. In his last competition, Güneş and his team worked on a recommender system project organized by Otto, a German e-commerce company.

Why did Otto start the competition? They wanted to improve the shopping experience for everyone involved. The problem? Predicting e-commerce clicks, cart additions, and orders using previous events in user sessions.

Customers want better recommendations while online retailers can increase their sales—it’s a win-win for both parties.

What about the data? Güneş notes that it was a huge json dump of 220 million rows.

 “When we convert this json to a dataframe, each row represents an event in a user session. It was hard to handle data with that size—so we had to use file formats like parquet and pickle.”

Güneş and his team generated three models for clicks, carts, and orders—and finished 28th out of 2,587 teams.

One of Davut’s biggest Kaggle challenges was a BNP Paribas Cardif Claims Management Competition. Banks know that claims management may require different levels of check before a claim can be approved and a payment can be made.

Kagglers were tasked with predicting the category of a claim based on features available early in the process—helping BNP Paribas Cardif accelerate its claims process and therefore provide a better service to its customers. The result? Davut’s team, Dexter’s Lab,  took first place among 2,920 teams. 

Davut, a Kaggle Master, has taken his academic, investment management, and Kaggle experience with him to co-found Aigoritma and thrive as the company’s Chief Data Scientist. 

Davut also enjoys the community and networking advantages of Kaggling. It’s where he met Ahmet Erdem. Ahmet, Data Science Director at Aigoritma, has had a journey quite similar to Davut’s. After a distinguished academic career, Ahmet worked as a Senior Data Scientist for NVIDIA, and became a top player on the Kaggle scene. 

On his way to becoming a Grandmaster, Ahmet was able to pull off the following Kaggle feats:

  • Sixth-Place Global Ranking
  • 21 Gold Medals
  • 25 Silver Medals
  • Four Bronze Medals

Ahmet attributes his success to properly validating models and effectively using deep learning for NLP, Computer Vision, and Forecasting competitions. For example, 

Corporación Favorita challenged the Kaggle community to build a model that more accurately forecasts product sales. 

Corporación Favorita were excited to see how machine learning could better ensure they have just enough of the right products at the right time. Ahmet and his team were excited to take home a gold medal for the competition. 

Ahmet’s favorite competition, however, was a PLAsTiCC Astronomical Classification. The challenge was classifying space objects based on their light curves. As the data was an unevenly distributed time-series—the problem was unique. By doing some creative feature engineering and utilizing both gradient boosting decision trees and neural networks—Ahmet took home his first solo gold and fourth place overall.

Now, Ahmet is Data Science Director at Aigoritma and is leveraging his past academic, data science, and Kaggle experience to develop projects for ML Studio.

Becoming a Kaggle Grandmaster

Güneş’s journey to the center has taken him from a hobby—to being ranked as high as 410 on Kaggle—to ML Studio, a platform which can be used in stages of similar solutions and can produce full-fledged competitive recommender system projects. 

Completing such projects is one reason why our next Kaggle champion, Davut Polat, co-founded Aigoritma. But where did his journey toward the center begin? Davut has earned a PhD from Istanbul Technical University and been the Head Data Scientist for a quantitative investment management company. Over the last decade, Davut has also been an avid Kaggler—reaching the following heights:

  • Kaggle Master with 10 gold medals, 12 silvers, and two bronzes
  • Discussion Expert—Highest Rank: 22
  • Competition Master—Highest Rank: 36

What Does It Take to Be a Grandmaster?

  • Win at least five gold medals in competitions
  • One of which needs to be a solo gold medal
  • Receive at least five gold and five silver medals in datasets
  • Win at least 15 gold medals in notebooks
  • Get at least 500 medals in discussions
  • 50 of these need to be gold medals

Would you like to start your journey and see how ML Studio can help your organization solve real-world problems with a ready-to-use End-to-End AI platform that doesn’t go off budget? 

Get in touch for a personal demo.

chevron-down linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram