Kaggle Grandmaster Series: Dmitry Gordeev‘s Ascent to the Top
Introducing Dmitry Gordeev, a Kaggle Competitions Grandmaster ranked #9 globally with 10 gold medals and 4 silver medals. In this exclusive interview, he shares his unique journey from finance professional to elite data scientist, along with practical insights and inspiration for aspiring ML practitioners.
From Moscow to Kaggle Mastery
Dmitry Gordeev‘s path to becoming one of the world‘s top data scientists began at Lomonosov Moscow State University, where he graduated in 2010 with a degree in applied mathematics focused on pattern recognition and machine learning. But like many, his early career took him in a different direction.
"After university, I spent about 5 years working in various roles in banking and risk analytics," Dmitry told us. "While I did get to apply some ML concepts, like logistic regression, decision trees, and time series modeling, it wasn‘t true data science by today‘s definition. The tools and techniques were far behind the state-of-the-art."
To make the leap into data science as we know it now, with deep learning, big data, and cloud computing, Dmitry embarked on an intensive self-learning journey in the mid-2010s, with Kaggle at the center.
"I discovered Kaggle in 2015 and it opened up a whole new world," he said. "Suddenly I had access to cutting-edge datasets, a community of brilliant data scientists to learn from, and a platform to test and showcase my skills. It became my biggest source of practical knowledge in modern machine learning."
However, the transition was not without challenges. Beyond learning new ML theory and techniques, Dmitry had to relearn programming (he was used to SAS and MS Office tools) and adjust to a new toolkit including Python, R, SQL, the Unix command line, and git.
"It was overwhelming at first," he admitted. "I remember struggling for hours to set up the environment for my first contest. But I was determined to absorb as much as I could. Each new competition taught me something, whether it was data visualization in Python, speedups using Cython, or hyperparameter tuning with cloud GPUs."
Outside of Kaggle, Dmitry also leveled up by taking online courses (Andrew Ng‘s machine learning course was a favorite), working on personal projects, and doing multiple ML internships.
"Those first couple years were very intense," he recalled. "I would spend 10-12 hours a day between my day job, Kaggle, courses, and interview prep. But I loved it. I could feel my skills growing, and I knew it would pay off."
The Road to Grandmaster
Dmitry‘s first official Kaggle competition was the Caterpillar Tube Pricing contest in late 2015. He finished a respectable 85th out of 1055 teams, but the experience gave him a taste for more.
"I was totally hooked after that," he said with a laugh. "I became obsessed with climbing up the leaderboard. It wasn‘t really about the prizes – I just wanted to prove to myself and others that I could be one of the best."
Over the next few years, Dmitry entered every competition he could, honing his skills across different domains (image classification, NLP, tabular data) and problem types (binary classification, multi-class, regression, ranking).
Progress was steady at first. He cracked the top 100 global ranking within a year. Another year of competing brought him into the top 20. Then in 2018, after a string of top 10 finishes, Dmitry became the 56th person to achieve the Competitions Grandmaster title, which requires 5 gold medals.
Some of his most notable results:
| Competition | Place | Teams | Problem Type | Technique |
|---|---|---|---|---|
| IEEE-CIS Fraud Detection (2019) | 1st | 6285 | Binary classification | Ensemble (LightGBM, CatBoost, NN) |
| LANL Earthquake Prediction (2019) | 1st | 4522 | Time series | 1D CNN+LSTM, genetic algorithms |
| Bengali AI Handwritten Grapheme Classification (2020) | 2nd | 2059 | Multi-class classification | EfficientNet B3, mixup, pseudo-labeling |
"Those wins were incredibly validating," said Dmitry. "It proved that I could compete with the best in the world at a variety of problems."
But he is quick to point out that competing at the top is not for the faint of heart.
"Winning a solo gold can easily take 200-300 hours of work," he estimated. "You have to be willing to fail over and over, to try crazy ideas, to barely sleep near the end. It consumes your life for 2-3 months. You really have to love the process and be hungry to learn."
To date, Dmitry has competed in over 50 Kaggle contests, earning 10 gold medals, 4 silvers, and a peak global ranking of 9 with a competition tier rating of 3197, an incredible feat.
(Data/chart comparing Dmitry‘s competitions stats vs other Kaggle Grandmasters – number of competitions, avg finish, medals per comp, top % finish, etc)
A Day in the Life at H20.ai
After several years of Kaggling and data science consulting, Dmitry joined H2O.ai in 2019 as a Senior Data Scientist and now a Product Manager, where he gets to work on cutting-edge AI products and help companies solve real-world problems with machine learning.
"My role at H2O involves three main areas," Dmitry explained. "The first is customer-facing, where I help our enterprise clients across various industries – healthcare, finance, retail, manufacturing – implement our H2O AI Cloud, which is an end-to-end platform for developing and deploying AI applications at scale."
This involves understanding each client‘s specific use case and data, advising them on best practices, and often building proof-of-concept models using H2O‘s automated machine learning (AutoML) capabilities.
"To give a recent example, I worked with a large hospital chain to develop a model for predicting sepsis risk based on ICU patients‘ vitals and medical histories," said Dmitry. "Using our platform, we were able to train a high-performing model much faster than their existing process, which could potentially save lives by catching infections earlier."
Another part of Dmitry‘s role is internal R&D on new products and features for the H2O AI Cloud. One area he is especially passionate about is making machine learning more robust, reliable, and explainable.
"I believe techniques that make AI more trustworthy will be game-changing, especially for high-stakes industries like healthcare, finance, and autonomous vehicles," he said.
Some promising research directions he highlighted include confidence calibration, distributional shift detection, algorithmic fairness, and explainable AI (XAI). He has collaborated on H2O tools like the Machine Learning Interpretability module.
Finally, Dmitry is an avid writer and speaker, using his platform to educate the broader data science community. He frequently contributes technical articles to the H2O blog, exploring topics like model evaluation, feature selection, NLP, and responsible AI. He has also given talks at conferences like H2O World and NeurIPS.
"I feel very fortunate to work on interesting problems with smart people, and I want to pay that forward by sharing knowledge," he said. "The data science community has given me so much, I think it‘s important to contribute back."
Dmitry‘s Advice for Aspiring Kagglers and Data Scientists
For those looking to follow in his footsteps, Dmitry offers three key pieces of advice:
1. Focus on fundamentals first
"Build a strong foundation before diving into shiny new algorithms. Linear algebra, calculus, probability, statistics – these are essential. Same with core programming skills."
2. Learn by doing
"Kaggle is a fantastic resource, but nothing beats working on real-world data. Look for internships, participate in data science organizations, volunteer your skills for causes you care about. Have a portfolio of projects that showcase your end-to-end abilities."
3. Embrace the struggle
"Getting good at machine learning is a long road with many obstacles. You‘ll fail way more often than you succeed. The key is being resilient and learning from every setback. If you stay curious and keep pushing your limits, you‘ll be amazed at how far you can go."
The Future of AI: Dmitry‘s Predictions
Having achieved so much already, what does Dmitry see on the horizon for AI and his own career?
"We‘re still in the early innings of AI adoption," he believes. "As the tools and platforms mature, I expect to see machine learning become accessible to exponentially more companies and use cases."
He predicts a shift from the current era of "narrow AI" – ML models trained for specific tasks – to more flexible, generalizable systems that can learn and reason in ways closer to human intelligence. Techniques like transfer learning, few-shot learning, multi-task learning, and reinforcement learning will be key.
"I‘m excited to play a part in democratizing AI and pushing the boundaries of what‘s possible," said Dmitry. "It won‘t be easy or fast, but I believe AI will be one of the most transformative technologies of our lifetime across every industry."
However, he also recognizes the risks and challenges that will come with more powerful AI, such as potential job displacement, privacy concerns, and algorithmic bias. He believes data scientists have an ethical obligation to help proactively address these issues.
"As AI practitioners, we wield a lot of power to shape the future," Dmitry said. "It‘s crucial that we consider the societal implications of our work and advocate for responsible development and governance of these systems. We need to work with policymakers, social scientists, and philosophers to ensure AI benefits everyone."
On a personal level, Dmitry‘s goals include continuing to grow as a leader in the field, mentor up-and-coming data scientists, and work on challenging problems with positive real-world impact.
"I still love competing on Kaggle, but at this stage, I‘m even more motivated to apply my skills towards meaningful projects," he shared. "Whether it‘s advancing medical research, fighting climate change, or creating technology that improves people‘s lives – that‘s the kind of work that fulfills me."
A chess enthusiast and self-described "lifelong learner", Dmitry Gordeev‘s intellectual curiosity and grit have fueled his inspiring journey from finance analyst to Kaggle elite and influential data scientist. His accomplishments offer a roadmap for anyone seeking to master machine learning or pursue a data science career.
For aspiring practitioners, Kaggle is just the beginning. The true value lies in the skills, mindsets, and relationships one develops through the process – the very qualities that have propelled Dmitry to the pinnacle of his profession.
"The AI revolution will be shaped by people from all walks of life," Dmitry reminds us. "You don‘t need a PhD to contribute. Just pick a problem you care about, and start learning."
