In this talk i'd like to give a technical deepdive of BigQuery. First, we'll visit the internals of the underlying storage system and other components that in the end make BigQuery SQL-like but not quite SQL and then we'll understand why this is such a good model for BigQuery as cloud product. This will help you as developer and end-user because you'll understand the strong points and limitations (quirks?) of BigQuery. After that we'll see how this plays out in real world scenario's and when you are impacted. Time permitting there'll be a section on cost optimization via partitioning and slots opt-in/opt-out policy.

Site Reliability Engineering is gaining more and more traction in software organisations all around the world. At bol.com there have been several initiatives mostly driven by individual enthusiasts over the last couple of year.

This year we started a dedicated SRE team.

This talk will cover what we've done to get to this point, what we've done since then and how the SRE team can help you innovate more effectively.

The cloud on the event horizon is about the how cloud can revolutionize our landscape in the near future. It has no direct bearing to our current journey and projects but it will show you how you can run a large scale service-based IT operation in a total different way than we do today. A resilient, scaling landscape where i will briefly touch on most cloud technologies. I'll illustrate this by showing parts of our existing customer journey so that there's not just hot air, but a meaningful relation to our current shop.

We haven’t invented time travel, so we can’t show you metrics from five years back that you didn’t store, but we have successfully created a set-up where, with little effort, you can start collecting metrics every tenth of a second, and store them for over a hundred years. And all this, with minimal effort or opex burden.

In the last year, we (Team Insights) have worked on developing a state-of-the-art metrics stack. However, this stack is limited to the constraints of our technical platform. Although convenient at times, a platform product doesn’t allow for the specific implementation that you might want. Perhaps you want your metrics measured each tenth of a second, or you would like to understand year-over-year black-friday performance, going five years back.

We’ll demo this brand-spanking new R2D2-based metrics setup, which will let you run your own metrics stack at minimal opex, but with the flexibility you need to get the correct insights.

Whenever we as software engineers read or review code, we need to understand what the code actually does. When writing code, we want to produce beautiful, but also readable code. Atoms of confusion are the smallest pieces of code that cause us to misunderstand what code will actually do. I am here to show you the results of my research on the effects these atoms have on understanding of code, and what perception CS students had on them. 7 atoms are shown to have a significant negative effect on understandability of the code snippets.

The best way to learn is from making mistakes. The second best way to learn is to look at other people making mistakes, and learn without getting hurt!

I will go through multiple mistakes that hampered our production environment. The focus of this talk is to tell fun and entertaining stories, and to then analyze what went wrong and how they could've been prevented. That way you can learn from our mistakes!

Software engineering is by nature a job for the creative and intellectually curious.

This means that we are intrinsically motivated to improve as much as possible around us, which can easily lead to being overwhelmed or having difficulty where to focus our efforts to make the most impact.

In this lightning talk I will share the one trick I used at bol.com which for the first time in my life made me feel like I truly was having the impact that should come with the promotions my managers kept throwing my way.

Not only that, it also significantly improved my focus, increased the impact of people around me and allowed me to share my privileges with others.

Machine learning has already made many professions obsolete. Data scientists keep automating business processes, which causes more and more people to lose their jobs. It seems like we are all threatened to be replaced by machines. Ironically, this includes the data scientists themselves. There are solid advancements in automatic machine learning itself. Is this something they should be afraid of? Are data scientists destroying their own future?
To answer this, we will first have to figure out what data science actually is. What can (and therefore will) be automated and what would likely to remain managed by humans at least in the foreseeable future? Next to these, we will find out whether robots can drink coffee, emotionally argue on what the KPI should be, massage your lovely stakeholders, and many more!

Are you fed up with all the design session meetings, marketplaces, review sessions?
Come experience mob programming in this fun collaborative session, solving a simple code challenge (kata).
We will learn to pair, agree even if we disagree, and create clean and sustainable features.

The kata will be implemented in Java using Visual Studio liveshare tool, thus the basic requirements are the use of Chrome as a browser and a Github account.

Experience the hive mind with us!

How do we verify our marketing budget is used effectively?
As bol.com, we are planning to spend around 90 million euros on digital marketing in 2020. The four main methods for determining marketing efficiency are: incrementality, media-mix modelling, lift studies and attribution. Here we focus on attribution, which provides a way of estimating how much of revenue is being driven by marketing channels and comparing their effectiveness. Currently we use the most widely used attribution model: last click interaction, which assigns all revenue to the last marketing channel which brought the customer to the website.
Our new system based on m2 data, will allow us to test basic attribution models such as: last click interaction, first click, linear etc. and to use the most advanced ML models such as Markov chain attribution, Shapley value and many more.
This will guarantee our digital marketing budget is being used in the most efficient way possible, helping us stay ahead of the competition.

Are you fed up with all the design session meetings, marketplaces, review sessions?
Come experience mob programming in this fun collaborative session, solving a simple code challenge (kata).
We will learn to pair, agree even if we disagree, and create clean and sustainable features.
And we will do it in Kotlin, cause here at bol.com we love it and we can still learn cool features from each other.

Experience the hive mind with us!

As a Data Scientist, I often need to quickly prototype ideas and showcase those projects to the business. Yet, Python scripts are not very visually appealing and powerpoint is not always suitable. And you also don't want to spend weeks setting up that throw-away, unmaintainable Flask application, just for the purpose of showcasing your idea.

Fortunately, there's a solution to all of these problems: Streamlit! Streamlit is an easy to use open-source app framework for Data Science written in Python which allows you to set up an interactive app in MINUTES. Streamlit takes away the pains of worrying about the front-end so you have enough time to worry about the data.

Give your demos and presentations superpowers by incorporating Streamlit into your data science toolkit. If you don't want to waste any more time building buggy and unmaintainable apps for the mere purpose of showcasing your project and want to go from zero to full-fledged demo in minutes, then don't miss this talk.

Are you fed up with all the design session meetings, marketplaces, review sessions?
Come experience mob programming in this fun collaborative session, solving a simple code challenge (kata).
We will learn to pair, agree even if we disagree, and create clean and sustainable features.
And we will do it in Kotlin, cause here at bol.com we love it and we can still learn cool features from each other.

Experience the hive mind with us!

"Your job should be your passion” has been repeatedly questioned by work/life-balance "gurus" with the argument that many people haven’t found their passion yet… But what if you have two passions? Can you find the balance then?
I personally gave up trying to find this holy grail. Instead of figuring out where to draw the line, don't draw one at all. Instead of trying to find 30 hours in a single day required to train on a professional level and work full time, do both at the same time!
In other words, a simple recipe for not being good at any of the two endeavors. Or is it? What if I tell you that you can learn more from training than from doing a PhD in mathematics*? What if the time spent in the office can actually make you climb harder than ever before! In this talk, you will learn about universal life (work?) lessons which will help you in the journey of becoming a more balanced human.
*Any possible formulas will be about nutrition, muscle growth, etc, but NOT about mathematics.

A few times last year, while implementing a solution, we were blocked by risks we did not foresee. Some of these risks could be solved, but others proved harder to overcome.

Based on last years’ lessons, we have been testing a new way of working. In this new way of working we go from idea to product, via POC, prototype and MVP. We iterate through the steps, implementing the idea more thoroughly at each step. After each step we try to prove the idea using experiments. This helps us to identify risks in an early stage and prevents surprises in later stages when already heavily invested in a solution.

In this talk I’d like to show how you can start iterating on ideas and prioritise the most feasible solutions. Helping you prevent surprises when there is no way back.

Humans communicate using much more than language. Our entire bodies are signposts of both conscious and unconscious signals that harbour multi-dimensional communication channels that are vastly more complex and efficient than we can imagine.

However, there is one on-line medium that comes close to this multi-dimensional communication power.

The Emoji.

This talk will dive the depths of opportunity inherent in this medium and will teach you to be a certified master of Emoji so you too can chat with ❤ and communicate with 💪!

Chaos engineering is the practice of doing experiments on your system to gain insights on how it will actually behave. In this talk I'll explain why I think we should do this more and challenge our concept of what makes up a system.