WE SPECIALISE IN FINDING FANTASTIC OPPORTUNITIES
FOR DIGITAL AND DATA SPECIALISTS WITH THE MOST INNOVATIVE BUSINESS ACROSS EUROPE AND THE USA.
[WATCH] 10 Key Points from our first AI in Data Science Meetup
We finally did it!
Hosting a meetup is something that myself and my team had been wanting to do for a while, but never actually got around to organising one.
We had an amazing response, with 153 people signing up for the event - filling the venue to 90% capacity!
Pretty good, right?
By the way, if you're looking for a Data Science role in Berlin, I probably have the perfect job for you.
You can also see how much you could earn in Berlin as a data scientist by downloading our Market Update.
We hosted our AI in Data Science meetup at Fusion Factory, right in the heart of Berlin:
Our second speaker was Sebastian Foucaud, who spoke about the Democratization of Machine Learning Products and the Impact on Business.
These are some of the points from his presentation that I found interesting:
1. Artificial Intelligence is not SciFi - it is HERE and NOW (6:13 in the video)
This is something I already knew, but having Sebastian talk about it really cemented it for me.
He said it's so important to technology businesses, that if the company hasn't got any views on it, then you are already too late.
We have it in our phones, homes, cars and even the grid power!
That just goes to show how critical it is already.
Imagine how vital it will be in five years?
2. We can build any AI-based product, but it doesn't make it a good idea (7:45 in the video)
The trend for consumer technology products at the moment is to advertise the AI & machine learning (it's almost become a buzzword at this point).
But does it make it better? Or useful?
Is it even necessary?
"AI for AI's sake is never the right way" - Sebastian Foucard
3. The model doesn't matter, it's the execution! (11:30 in the video)
This just goes to show that a data scientist really can work with what they are given (most of the time).
Some ways might be easier, some might make it more difficult.
If you have the right technology and talents, it should work!
And there are tons of open-source software available, and there is a huge amount of talented data scientists.
I only work the Berlin market and have connections with thousands of them!
My experience is that:
4. You need to know your customer before they visit your website (or app) (24:56 in the video)
User experience is vital for any business, app or website.
It can be the difference between a paying and returning customer, to a customer who never interacts with your brand again.
Sebastian made a great point about when he was looking to buy a car.
The app wanted to know what make of car he would like to buy.
He had no idea (and he didn't care).
He wasn't looking for a particular make of car, he was looking for a car that would suit his needs.
So, that means:
5. Personalised recommendations are the way to go (28:28 in the video)
The last point which stood out to me about Sebastian's presentation was how he maximised the profits from users who used two different platforms.
There are two different platforms, one for buying houses, and one for buying cars (only 5% of users are on both).
Somebody viewing a farmhouse was shown an advert for a truck.
See the connection?
Information was plugged in from other users to achieve this recommendation.
These are just a few of the interesting points from Sebastian's presentation, you can see the rest by watching the full video here:
Our other speaker on the night was Luba Weissman, who spoke about the Potential of Behavioral Data in Credit Scoring.
Luba has been a data scientist for the last 10 years now, and for a big part of her career, she's been in the finance industry.
Here are 5 points from her presentation that I found interesting:
1. AI considers what linear models could not (3:28 in the video)
Linear credit scoring models in the past have been limited to what they can analyse and make decisions on.
This affects people from all walks of life and around the world!
As Luba explains, people without payments/credit history, immigrants and the younger generations would have been denied or not even considered before.
AI can look past these factors and delve deeper into making the correct decision.
As if that's not enough:
2. Regulators can slow down these processes (4:55 in the video)
I suppose this is the same in any industry, really.
But this is especially important as the data of huge numbers of people are at risk here.
But that's just part of the story:
3. Reality isn't linear (5:36 in the video)
This one makes a lot of sense to me.
Humans aren't predictable, and regular techniques hit a limit on how often they are right or accurate.
AI tries to alleviate this by using non-linear techniques.
Think about it:
4. Customer behaviour isn't only limited to Facebook (6:58 in the video)
I knew that customer spending behaviour was tracked for credit scoring, but I had no idea just how much was tracked.
Using your card in a casino (or at an ATM close by) has an impact.
Change your smartphone? It'll have an effect.
Call customer service a lot? That'll have an effect too.
Crazy stuff, right?
5. We GIVE companies this data
As soon as I heard what has an impact on creditworthiness, my next logical question was.
How do they get it?
I'm sure yours was too.
Turns out, we give it to them.
When we sign up on websites or use the apps on our phone.
Social networks have a big part to play as well.
Luba's presentation had a lot of interesting insights into credit scoring and what part AI has to play in it.
You can watch the full presentation here: