Hackathons are the coolest. Where else do you get to thrive amongst your peers in an intensely competitive environment, where you can have fun, build a useful product, learn, meet people, and get exposure. In the past years, data science hackathons have also become increasingly popular amongst data scientists and developers, because they like seeing people actually doing AI, natural Language Processing, or deep learning, instead of just reading about them.
There are loads of articles out there on how to win these things of course. Basically, be smart on your feat, adapt, and present well. But what nobody really talks about is what you should do to bomb a data hackathon. From our experience, this is it.
Let’s start off with a few pieces of advice that apply to life, in general, but also, data science hackathons.
You must absolutely work through the night. If you’ve got to pump Redbull through your veins to make it, do it.
Being social is a waste of the precious time you could be spending on coding, so get on it. And even if your teammates did have something valuable to teach you, it’s too risky.
Show them all how much fun you’re having by posting lots of pictures of the free pizza you’re eating next to your Jupyter notebook #g33klife (or whatever it is the kids are saying). Multitasking is key. Focus is for the weak.
And leave the debugging to everyone else. Why? Because passion always pays off ❤ Also, it’s obviously better if you’re really good at one thing and one thing only, so you won’t be asked to do the rest and won’t be able to help your teammates out, and again, you’ll save lots of time. Also, commenting code is a waste of time spent typing, and indentation is for losers.
We all know the Ballmer Peak. Is it real? Is it fake? How much alcohol can you drink before it affects your efficiency? You should definitely find out if you’re serious about winning this thing.
That’s it for your life lessons of the day. Coming up next are a few tips specific to how to rock a data hackathon.
There’s no better way to start a hackathon than by coming in with a completely empty brain, open to all the ideas that other people spent the night working on to get ahead.
Only work with people who think exactly like you and have the same skills to avoid all friction (see tip 8)
”They bother people.” The only thing they’re good at is writing convincing Powerpoint presentations and telling stories. Nothing you’re going to need. When it comes time to present what you did to the judges, your code will do the talking. Alone. And what if your judges are business people and project managers? Well, they obviously should have learned how to code.
You have to be sure you have THE perfect tool for absolutely every step of your process. That means one tool for data preparation. Another for machine learning, exploratory data visualization, web apps, and so on. Then, find a tool that the people you aren’t talking to (see tip 2) can pretend to collaborate with by merging all of their work into one project. So yeah, good luck with that.
So do your research and benchmark all the different tools you can find, with documentation, so you’re absolutely sure that task management app you chose is THE BEST one out there.
The best part of any hackathon is the final countdown, scrambling through your files for that final version of the script because you’re not sure which is the right final dataset between : data-prepared-cleaned-deployed-v36 and data-prepared-deployed-enriched_v22 And realizing that the difference between these is that the person in charge of cleaning and enriching didn’t start with the same versions. You’re a code warrior, enjoy it!
People at hackathons don’t care that you've made something remotely useful and practical that speaks to non techy-people and therefore has a chance of having an impact. They just want to see how far you can push the technology. If you can’t even explain your results, that’s a good sign.
I know there are lots of tools out there that can make it much faster to do your data cleaning than just coding everything up in SQL, Python or R. But you can’t chose the easy way out. If you’re a Python guy do as much Python as possible; And bootstrapping with someone else’s code that you found online is cheating. Plus, the only person you can trust is yourself.
And if there’s a platform out there that makes it easy to connect to data sources (like pretty much any source including Hadoop and Spark), helps you visualise your data to power through the exploratory phase, with accelerators that make your data wrangling faster and less repetitive, that also allows you to bring in your code and share code samples with your team, and gives you an environment to iterate on your workflow and build web apps interactively, and collaboratively, run away. A hackathon should be hard, a bit lonely, and often frustrating. You’ve got to feel the burn. So if you ever find a tool that makes it easi(er), just run.
All in all, our biggest tip to you: don’t use Dataiku Data Science Studio for your next data hackathon. Whatever you do, don’t click on :
Please fill out the form below to receive the success story by email:
How can we come back to you ?