Whitepaper – The Analytics Translator
The Analytics Translator is the liaison between senior management, the business, and data experts. On some days, you are the gatekeeper of the project funnel, brainstorm ideas with executives, and work with the data experts to groom the backlog of viable ideas.
When other technically inclined friends learn GoDataDriven is in the "data" space — science, engineering, analytics translators — I often get the "look".
In the "look" you can read what mainstream news reports about the "data" space. Scandals about the US election, manipulation of the public before the Brexit referendum, VPN apps that spy on you to target ads even better, consultancy companies that advises pharmaceutical giants how to sell more addictive drugs.
Is it true, as Jeffrey Hammerbacher said, that
“The best minds of my generation are thinking about how to make people click ads.”
No! We certainly are not: the majority of our projects is focused on doing good things, things that don’t require my personal data to be accomplished and if they require it, they’re not using it to do creepy things. Instead we create models to accomplish goals we are proud of when speaking with friends and family!
What models then?
Reducing bread waste
For one of our clients, we implemented a system to reduce bread waste by optimizing how much is delivered every day.
Bread is a particular product when you want to predict how much should be delivered tomorrow, because — at least in the Netherlands — you cannot keep it on the shelves for more than a single
day. This rules out almost all shelves-replenishment systems out there.
That’s where GoDataDriven came in. Using historical unit sales, weather, holidays, and more data sources, we were able to greatly reduce the amount of bread that is wasted, every day. Yay!
Reducing flower waste
You don’t hear often about flower waste, do you? Well, we heard about it. When growers want to sell their flowers at the largest flower auction system in the world, they need to provide a picture of their flowers. If the picture is not good enough, nobody in the auction will buy the flowers and they will be discarded.
To help the growers, we built an image recognition system that assesses the quality of the picture taken. If the picture can be improved, we offer actionable advice to the growers: the can retake the picture, sell the product, and no flower gets wasted. Yay!
Avoiding trains stopping in the middle of nowhere
If you ever rode a train in a densely populated area such as the Netherlands, you know that a train that breaks down in the middle of nowhere can be a disaster. Especially if out of service, other trains have to stop, deviate to avoid the broken train, etc.
Two of the most common causes of train breakage in the Netherlands were bearings and compressors. When the first would fail, the train had to stop immediately (alerted by sensors along the railway). When the second would break, the brakes would shut and the train would quickly come to an halt.
GoDataDriven helped the Dutch railway to build the data platform and the algorithm that would predict — between 60 and 15 days in advance! — when these components where going to malfunction.
We therefore implemented a system that would warn the technicians that a particular bearing was breaking down or that the compressor air system had a leakage. Safer and more timely trains for everyone. Yay!
Improving traffic light efficiency
Every time I go back to Italy, my patience quickly dries up at every traffic light. Why? They didn’t change since I left the country more than 10 years ago.
The answer lies back here in the Netherlands: practically every traffic light here is smart! With sensors inside the adjacent roads, the traffic lights will respond to the number of cars and will not make anyone wait unnecessarily.
Together with the company responsible to create the models for the smart semaphore, we set up a data platform to collect and analyze all the data coming from the sensor to determine when the traffic light models were underperforming.
Now I lose my patience even more quickly when I visit a country without these sensors. Yay!
If you come from the Netherlands, I probably don’t have to convince you that the lack of good beers is a national security question.
That’s why we teamed up with a brewery to improve the brewing process by creating a model that, taking into account the beer color through sensors, suggests how to produce the tastiest brew. Yay!
Medicx is probably our best kept secret. Part of GoDataDriven and with the mission of making medical technology smarter with the use of data and artificial intelligence, Medicx uses medical
data to create machine learning models.
One example is CardioAI, a project applying deep learning to MRI data for automated inpainting of the left- and right- heart chambers and the heart muscle. Automating this task, especially the extraction of characteristic metrics of the heart function, save tons of time to health
This is just a selection of the work we are all proud to do. I hope the cases presented will convince you that you don’t have to optimize my clicks to be called a data scientist.
Oh, did I forget to tell you that we sponsored CardioAI ourselves, because we really believe data can be used to make the world a better place?
Want to be involved in creating fair algorithms in your organization?
Join us for the Data Science for Product Owners course that will dive into ethics and fairness in AI as one of the course topics.
If you want more rambling throughout the month, follow me on Twitter: I’m gglanzani there!