Between November 13th and 15th, 2022, Kagglers from all over the world came together in Paris to network, compete, and learn from each other.
What are Kaggle Days?
Kaggle is a platform mainly known for its Data Science competitions, and Kaggle Days is a series of worldwide events for Kagglers organized by Kaggle fans. The non-profit event series was created by Maria Parysz, founder of LogicAI, in 2018 and has already had six events and more than 90 meetups.
Kaggle Days is a series of worldwide events for Kagglers organized by Kaggle fans
To participate in the event, you had to apply for a seat up front. Applicants were approved based on their experience in data science and activity on Kaggle according to the Kaggle Days Paris 2022 Terms and Conditions.
This edition of the event series was organized by LogicAI in collaboration with Kaggle and LVMH, which specializes in luxury goods, including Louis Vuitton and Moët Hennessy.
The two-day event was held in Paris with the following agenda:
- Evening before the event: city tour and evening get-together for early arrivals
- Day 1: 9.5 hours long Data Science competition at the Google offices in Paris
- Day 2: Data science presentations and workshops by Kagglers and LVMH in parallel to the LVMH Data Summit at Pavillon Cambon.
Networking with Kagglers and other Data Scientists
If you are intimidated to participate alone in a Kaggle Days event, don’t worry: My impression was that most participants didn’t know any of the other participants before.
Aside from the Data Science talk, the Kaggle Days event’s main attraction is to network with fellow Kagglers – and the event organizers understand this and actively help you make it a successful networking event.
The event organizers understand this and actively help you make it a successful networking event.
Before the event starts, you are invited to a Slack channel to connect with other participants. On the day before the event, there was a small get-together where you had the chance to meet participants in a small group over snacks and drinks. There are plenty of coffee breaks during the event, and every event day finishes with a get-together over drinks.
Luckily, the Kaggle community is not only warm and welcoming online but in person as well. Within 2 minutes of entering the event location on the first day, I met a fellow Kaggler who became my teammate for that day’s competition.
The way we teamed up was not an exception. The main room for the competition was filled with large round tables that could fit about eight people. My teammate and I sat down at a table with a group of Kagglers engaging in a friendly chat. When we asked them, “How do you know your teammates?” their response sounded familiar: “Oh, we just met.”
“How do you know your teammates?” – “Oh, we just met.”
But what if you don’t find a teammate before the competition? The organizers of the Kaggle Days event have got you covered for that as well! Before the competition started, Kagglers who wanted to join a team could gather in a dedicated spot. That is how we expanded our team from a duo to a trio.
How to Win a RecSys Competition
The first day of Kaggle Days Paris 2022 started early for most Kagglers. At 8:00 AM, the event was opened and kicked off with a 9.5 hours-long Data Science competition in the style of a Kaggle competition. The competition aimed to build a recommender system that could predict the correct search relevance of products for a searched query. Ironically, “How to win a Recsys competition” was the title of Gilberto Titericz’s presentation on the second day. This probably would have helped a lot of participants in this competition.
After the competition topic was announced, the competitors got to work immediately. Although I assume most participants just wanted to have fun, you could feel the pressure in the room: 9.5 hours to come up with a solution is not that long.
The first hour flew by in a blink of an eye: Start downloading the data – the WiFi is not working… –Oh, there is a second WiFi! Try that one – explore the data – review the problem statement – set up a team with your teammates on Kaggle – set up a collaboration infrastructure…
Once everyone in the team was up to speed, it was time to discuss ideas. We aligned with our mentor (which I will get to in a second) and brainstormed some possible approaches.
In my team, we decided to split up the work: While one built a submission baseline, including validation, and made a first sample submission, the other two got to work and tried different approaches. “Lunch is served!” – “What? It’s lunchtime already?”
“Lunch is served!” – “What? It’s lunchtime already?”
In the afternoon, we were able to make the first submissions. “What does the score on the public leaderboard say?” Make some changes and submit again. Fix some bugs and submit again. Rinse and repeat until you are out of ideas.
This was the part of the competition when the resignation started to settle in. Usually, this would be the part where you would go for a walk or go to sleep to start over with a fresh mind. But in a hackathon, that’s not possible. Tick, tock – only one hour left.
Usually, this would be the part where you would go for a walk or go to sleep to start over with a fresh mind. But in a hackathon, that’s not possible. Tick, tock – only one hour left.
In this competition, 85 Kagglers and 55 LVMH data scientists competed in teams of up to three people. While this competition did not award Kaggle medals, the winners received some nice prizes, such as Louis Vuitton speakers, Sephora gift cards, and Moët & Chandon champagne.
So, how do you win a RecSys competition? You can read all about it in the 1st Place Solution, including the code.
What I found interesting was that the simple baseline of products sorted by popularity was already challenging to beat. Also, some teams were quite successful without using any Machine Learning approaches.
I’ll add more resources as they get published by the speakers or event organizers.
What It Is Like Having a Kaggle Grandmaster as a Mentor
While you were required to bring your laptop, the event organizers provided WiFi, snacks, and drinks. And on top of that, each team was assigned a Kaggle Grandmaster as a mentor. One of the most exciting experiences during Kaggle Days is that you will have a Kaggle Grandmaster as a mentor to guide you through the competition. While this mentorship experience will be unique to your team and your mentor, what you make of this opportunity is in your hands.
While this mentorship experience will be unique to your team and your mentor, what you make of this opportunity is in your hands.
At the beginning of the competition, each team got to draw a mentor from a deck of cards facing down. You could choose the mentors from the Kaggle Grandmasters, the speakers of the event, and Kaggle’s Walter Reade himself – pretty much a deck of aces if you ask me.
Again, this experience will be unique to you and up to you what you make of it. While some might not have talked much with their mentors, other teams worked closely with theirs.
In retrospect, my team and I could have probably taken more advantage of this opportunity and asked our mentor more questions. Nonetheless, I am happy with our experience and have learned quite a bit from the mentoring in this competition.
Here are my key takeaways:
- To get started, check out approaches to similar competitions and build from there.
- Start with a simple approach. Once you have submitted something, try to improve your solution step by step.
Anyways, if you have to remember one thing from this article, it would be to take advantage of this rare opportunity. Ask them how to approach the competition, deal with issues, or what you should do when you are stuck.
Day 2: Presentations and Workshops
Macarons for breakfast and a day filled with interesting talks and workshops? Yes, I’ll take both. Thank you very much!
The second day of Kaggle Days Paris 2022 took place at Pavillon Cambon and was organized together with LVMH. The location was stunning, and as you can expect, as LVMH specializes in luxury goods, I felt underdressed in my Kaggle t-shirt and sneakers among the stylish LVMH folk.
The day was split into three parallel streams of talks and workshops:
- Business Stream: A stream of shorter 30-minute presentations by LVMH
- Data Science Stream: A stream of longer 45 minutes or more presentations by mainly Kaggle Grandmasters
- Brainstorming sessions: A stream of 1.5 hours long workshops and brainstorming sessions for smaller groups of about 20 people
Presentations from the Data Science Stream
The talks by the Kaggle Grandmasters were both divers in topics and also highly insightful. You can check out the presentations below:
I’ll add more resources as they get published by the speakers or event organizers.
- Jean-Francois Puget: “Machine Learning for Fraud Detection”
- Philipp Singer & Yauhen Babakhin: “Practical tips for deep transfer learning”
- Ahmet Erdem: “Using Machine Learning to discover the factors affecting product return”
- Andrada Olteanu: “Data meets design”
- Gilberto Titericz: “How to win a Recsys competition”
- Bojan Tunguz: “Automating Business Tasks With Generative AI – A Survey Of Recent Progress”
- Theo Viel: “Item matching”
As you can guess, I wasn’t able to see all presentations due to the parallel streams, but for the ones I saw, I can say that the presentations were of high quality and very insightful. E.g. I especially liked the interactive presentation by Philipp Singer & Yauhen Babakhin, during which the audience got to vote on the hyper-parameter values for tuning a baseline model via slido.
Workshops
- Dmitry Gordeev: “Creating agents for simulation competitions”
- Darragh Hanley: “Transformers for Document Information Extraction”
Brainstorming Sessions
The brainstorming sessions were held in smaller groups of about 20 people. You had to register before the event for these limited seats. These brainstorming sessions aimed to develop an idea for an LVMH business problem.
We were split into groups of four to five, with at least one LVMH business person, one LVMH data scientist, and one Kaggler. At the end of the brainstorming sessions, each team had 2.5 minutes to pitch their idea and another minute to answer questions from the judging panel. The team with the best pitch received some prizes from LVMH.
Let me tell you this: It was challenging to discuss the problem statement and develop a presentable solution in just about an hour.
Conclusion: Returning Home with More Than SWAG
All in all, the Kaggle Days Paris 2022 event was an amazing two days spent among fellow Kagglers. We all return home with something more valuable than what was in the goodie bag. – No, I’m not talking about the Louis Vuitton speakers. We got to connect with other Kagglers and learn from Kaggle Grandmasters.
So, would I participate in another Kaggle Days event, and would I recommend it? – Yes, and yes!
So, would I participate in another Kaggle Days event, and would I recommend it? – Yes, and yes!
You get to meet and connect with other Kagglers and learn directly from Kaggle Grandmasters. If you are an active member of the Kaggle community, I recommend participating in a future edition of Kaggle Days. To ensure you don’t miss the next event, you can subscribe to the newsletters or follow the Kaggle Days LinkedIn page.
The Kaggle Days event organizers, LogicAI, have kindly – on my request – permitted me to use their photos for this article. This is neither an advertisement nor a sponsored post. All opinions are my own.
This blog was originally published on Towards Data Science on Nov 22, 2022 and moved to this site on Feb 1, 2026.