the exciting applications of Data Science

View fullsize

(by Eric)

As a Data Science intern at Johnson and Johnson, I've had the privilege to dive right into two full projects. By the end of the summer I'm expecting to complete a third. My experience here has been nothing short of enriching, interesting, and challenging.

At the moment, I am pair programming with another intern to tackle a machine learning problem. We are coding an algorithm that takes raw customer complaint data in the form of Excel files and assigns to each row of text a one word category that best summarizes the description. This effort involves the use of Excel to compile and clean the data, Python scripts to implement and test various machine learning algorithms, and Tableau to visualize the results and present them in a digestible format for our business users. I can definitely say that the Statistics and Computer Science courses I've taken at Columbia are proving their worth now!

View fullsize

Writing code very intensely with my project partner Ryan

What I've really enjoyed about working in my Data Science group is the level of trust and freedom that we are given to carry out their work – no one's looking over your shoulder to check on your progress or make sure you finish things on time. At the same time, there isn't complete chaos or lack of structure; we hold project meetings regularly to help align efforts and ensure that each contributor is given the resources he or she needs to deal with obstacles and problems as they arise.

What's more, we interns not only carry out projects from start to finish, but we also get to own the entire result by presenting proof-of-concept demos and updates to the business users (many of whom are J&J department directors and managers). For us, this means that beyond coding and analyzing data, we're challenged to think about the broader impact our work has and find ways to communicate very complex ideas to a non-technical audience.

“By showing our capabilities and the power of Data Science to solve important business problems, we get to shape the image and future success of our department.”

View fullsize

The inside of the working space

As was mentioned in the previous Blind Applying post, Data Science is a relatively new concept to many people in the company, so this gives us interns a unique opportunity. By showing others in the company our capabilities and the power of Data Science to solve important business problems, we get to shape the image and future success of our department. Pretty exciting to say the least!

There's a lot more I have to say about my summer experience, but I think I'll wrap up my first post here. Stay tuned for more about my projects, where I work, and what life is like at J&J.

View fullsize

The window view right by where I sit

Time flies… and things get done

(by Gonçalo Guerreiro)

I started my internship at ABB in the beginning of March and it has been an incredible experience so far. From the first steps in the office, to the opportunities of wandering around Krakow, a lot has happened and, as my project moves forward, I realised I haven’t shared much about it yet.

Since I joined ABB, I have been working to develop a proof-of-concept for a “new” type of controller. This means I am building a simple version of a controller to test the feasibility of a new approach. If successful, the technology can then be improved, scaled, and may be applied in future products.

What can it be used for?

Controllers are very common in today’s technology: industrial processes, automotive applications, aerospace industry, among others. They are capable of monitoring a system’s operation and control its future behaviour. The controller I am working with could end up being used in almost any application but, for now, the plan is to test it in a water-pumping system and in a servomotor to assess its efficacy.

New?

Control has been around for quite some time and it’s widely used. The novelty in my project comes from the usage of parallel processing in a predictive controller. Predictive controllers calculate the input for a certain system by comparing its actual state to a programmed desirable behaviour in order to match them. Predictive control problems can easily become very complex and difficult to handle. However, it is very likely they could be solved more efficiently with the use of multiple processors rather than a single processing unit (the usual approach). This possibility has motivated my research and I have been exploring it for the past two months.

How am I doing it?

In order to develop the project I have been using as a main tool, one FPGA (Field-Programmable Gate Array) in which I implemented a multiprocessor architecture. I am programming the system using the C language. While I was not a stranger to some of the tools / concepts I am using, the amount of learning  has been incredible.

The status

I am happy with the developments so far. I already have a basic working version of the controller and the next steps will be its improvement and testing in a real situation.

That was the sum-up of my project so far. I hope it has some appeal for the engineers out there and that I didn’t geekify it too much for the others!

Some other updates: I have just received a video camera from ABB to do some video diaries from my experience during the internship and next weekend I will be travelling to Zurich for a week of workshops at the global headquarters together with global trainees from all over the world. I will keep you in the loop!