Data Science Retreat

Data science retreat is a unique cocktail of an advanced degree with a professional focus. Students learn from Europes top professors and professionals for theory, but unlike academia, classes focus on how to apply this knowledge and write code to solve real business case studies. After the curriculum, you're responsible for presenting a portfolio project of your choice, a fruit of self study and investigation, to a room full of Berlins CTOs and chief Data Scientists. Unlike many similar programs, the program fosters a highly individualized education, optimizing how data science could be best used by each students set of skills, rather than simply molding an armada of "code monkeys" capable of merely blankly regurgitating a pre-fix menu of algorithms or tasks. My only lamentation, through no fault of the retreat itself, is that while job assistance in Europe is increadibly strong, it's Eurocentric nature is not ideal for those returning to the US. On the other hand, its education and training are unparalleled. 

So I didn't have the patience or time to go back to uni and it was getting increasingly lonely doing online courses, and topics such as Apache Spark are simply not available in online courses. I now see why, it is quite fiddly to get it to run on your machine and so the instructors from the two sections we had on it (RDDs, dataframes and real-time/streaming) went through the set-up process. There were many great instructors in this course, again intensive training in NumPy and Pandas and some detail on the tricks within scikit-learn. The neural networks component was one thing I saw in this course for the first time, and am happy to have developed some proficiency with it. There was also Python environment management I've been dying to learn for quite a while since I damaged my system python once and had to reinstall everything. It is the proper way to do projects. There were a number of topics introduced with less detail such as big O notation and microservices which are important to know about and apply.

There were a lot of fantastic networking opportunities around DSR, with several alumni and succesful industry contacts sometimes joining for lunch or class.

I was a bit disappointed that in Germany no aspect of the course was in German and there were no German participants, but several of the instructors as well as the managing director are German. Still, it was a great group of people socially and professionally and I look forward to staying in touch.

Also, the process of converging on a project idea was more painful and less smooth than it could have been. It needs to be described in writing, it's like they think you can read their minds regarding what they are expecting. Hopefully they will fix that by the time you sign up for a course, and it is in any case a great opportunity to add something impressive to your online portfolio.

It is a very intense 3 months which was what I was looking for.

A refreshing experience that pushes you to explore further data science and create an ambitious portfolio project. You will undoubtly get a job at the end of the course.

During the bootcamp I personally felt the switch from learner to skilled practitioner, and it gave me great confidence in my abilities as a data scientist. I enjoyed the informal atmosphere and collaboration with my peer learners.

The pros:

- Small batches (8 to 10 participants), good selection of participants (expected strong background in ML)

- DSR covers what online courses don't cover. Assuming you come well prepared (already know python and machine learning), it will boost your skills and make you employable

- Most teachers are extremely skilled and really know their area. Some might not be able to follow though, so expect to spend hours after class to make sure you absorbed all knowledge. Teachers are available outside of class hours to answer you questions.

The cons:

- DSR claims to use the meerkat method but sometimes it sounds more like an excuse for lack of ability to teach of some of the teachers

- Still managed in a very artisanal manner, lacks consistency (e.g. teachers work independently and tend to cover the same topics, the sequence of topics is not always logical) 

- Few teachers not up to the expectations for a course of that price

Pros - 
    -  I find the screening and selection process quite intensive. So, everyone attending data science retreat are more or less in an intermediate or advanced level. So, you learn a lot from your batch mates.
    - Lecturers are experts in their respective field
    - The regularly updated syllabus me
    - Smaller batch size, in my batch it was 7
    - Attracting practising data scientists to sharpen their skills ( had a batchmate, who was working as a data scientist in a company)
    - Supportive staff, spacious place to work and discuss

Not sure pro or con-
    - Pressure to come up with data product with value for a project portfolio.  

Cons -
   - Meerkat method? am still thinking whether it was used or not

The Data Science Retreat is one of the few full-time data science courses in Europe. The 3-months intensive program is... intense. The courses are taught by senior data scientists from industry or from professors in the field. They are taught in blocks of 1-4 days and the topics cover a wide range.

Where the order of the courses is sometimes a bit confusing, and not all lecturers are of equal quality (both in terms of depth of knowledge and in terms of how they teach), the retreat does give a very good and broad overview over the important field of data science, AI and ML.

During the 3 months course, all students are required to do a project that results in a public presentation in the final week where industry is invited (networking for a job!). The 3 months are therefore split up into about 50% lectures and 50% project work. Mentors/ Tutors are assigned for the project work for weekly one-on-one meetings; this is also very helpful. 

The batch-size is small (6-8 participants) and there is therefore plenty of time to interact with the lecturers, discuss and ask questions. 

Things might appear chaotic at times, but overall it's a very good experience. 

Furthermore, Berlin is a fun city and very reasonably priced. 

Pros

• I attended a recent batch which started in September 2016
• High standards for acceptance was great - meant we could move fast and learn a lot not only from teachers and mentors but also from other participants
• Great benefit is lots of teachers, each with their own speciality - you get cutting edge advice and material about things the teachers are using themselves in their full-time jobs
• High quality of mentors and teachers and support for the portfolio project
• Great curriculum and decent job support

Con

• Not for everyone - Requires a lot of hard work to get the most of it, but a massive plus if you are very motivated  - the course moves at a pretty decent pace and you are not spoon-fed how to do everything, it requires working hard on exercises and a hands-on approach. However, this means you cover a lot of ground and learn very quickly how to figure things out as fast as possible, as you would in a real job

I was a participant of a batch in 2016.
My overall expectation was to become an expert in predictive modeling.
But there my expectation was not clearly met, because I found out.
However, the hands on experience offered to me was really useful.
What we covered was Random Forest, SVM, Deep Learning.
In the end, there is no free lunch. There is no shortcut to master all the topics.
But all in all there is a lot of pointers to various directions.
So the total experience is to extend my horizons, and it amazingly did.
I now know where to look for further information, what is possible.
You get to know which cutting edge technology is being used by professionals and how they apply it.
It was a multiplication of knowledge in general and I am satisfied with the outcome.
I see the portfolio project as a very useful tool to associate what you learn with a application context. However, it will improvement the experience a lot more if the students don't need to concern with data engineering.