I have recently graduated from the Metis Data Science Bootcamp (Singapore, Batch 5), and enrolling in the Bootcamp might have been one of the best decisions that I have ever made in my life. Out of the mandatory 5 projects that I have completed, all have been published on Towards Data Science (TDS), and 2 have been featured on its social media. Most importantly, however, I managed to land myself two job offers as Data Scientist even before the Bootcamp concluded. Pardon me, but being a beneficiary, therefore, I wish to share with aspiring data scientists on the Bootcamp, the pros and cons of it, and how to leverage on it to derive the maximum benefits.
What is Metis Data Science Bootcamp?
In summary, Metis Data Science Bootcamp is an accredited 12-weeks project-based and immersive apprenticeship in full-stack data science. The Bootcamp is very strong in breadth, covering a wide range of data science domains, from communication and design, exploratory data analysis all the way to big data and neural networks. In bullet-points, let me describe a list of main topics that I have learnt:
- Communication and design of projects/presentation
- Career Preparation (Resume-building, Networking, Interviews)
- Exploratory Data Analysis
- Data Acquisition by Web Scraping
- Model Building (Regularization, Cross-Validation, Feature Selection/Engineering)
- Supervised Machine Learning (Regression Algorithms)
- Supervised Machine Learning (Classification Algorithms)
- Unsupervised Learning – Clustering, SVD, PCA
- Natural Language Processing
- Recommender Systems
- Neural Networks (ANN, CNN, RNN)
However, even the long list of topics above is not exhaustive of all that the Bootcamp covers! As such, to compress all these data science knowledge into a 12-week course, one can imagine the extent of intensiveness of the Bootcamp. Notably, my instructor told us that it is impossible to master all that is taught within the Bootcamp, and in fact he took around 2 years to be able to assimilate these knowledge.
The Structure of the Bootcamp
The Bootcamp requires a full-time commitment (Monday-Friday) for 12 weeks, with classes starting at 9am and ending around 2pm on average.
Every day, classes start with Pair Programming, where everyone pair up with a buddy to work through a programming or data science problem. The Pair Programming started off manageable for me in Week 1 and became progressively more difficult. Somewhere around Week 6, after Recursive Functions and KNN implementation, I started to become lost and so did my other peers. Pair Programming problems by Metis were truly difficult.
After Pair Programming, the instructor typically goes through slides and Jupyter Notebooks on data science concepts and machine learning tools, that were usually catered towards completion of the subsequent project. Because these lessons usually end around 12–2pm, the lessons were typically short and intensive. Hence, it is paramount to open all ears, and give maximum attention.
Although the lessons end around 2pm typically, I caution against spending the rest of the day in some sort of part-time work or commitments. Apart from resting the brain from the morning’s whirlwind tour of data science concepts, personal time should be spent revising the concepts and reading ahead to deepen understanding. Remember that I mention that the Bootcamp is strong in breadth? The Bootcamp is weak in depth, perhaps intentionally. Hence a lot of self-exploration (using Google, in many instances) is required on our part to pick up where that is left in the lesson. More importantly, these extra time are given to us to work on our projects!
In total there are 5 data science projects to be completed as part of the Bootcamp. At the end of it, we are required to give a presentation to the class that is limited to 4 to 5 minutes. Also, we are strongly encouraged to create a blog post of our projects. At the end of project 5 (the last day – Career Day), we also give a final presentation to a panel of prospective employers. Below are the chronology and descriptions of the projects in the Bootcamp:
Week 1 – Project 1: Exploratory Data Analysis
In Project 1, everyone worked on the same project – MTA Turnstile Analysis. We were given a fictitious business problem and were tasked to apply exploratory data analysis to advise on solutions to the business problem of social engagements. We made use of the lessons taught on Pandas, Numpy, Matplotlib and Seaborn for our project. Below is the project that I have done:
MTA Turnstile Traffic Analysis to Optimize Street Engagements
Week 2 to 3 – Project 2: Web Scraping and Regression
In Project 2, we experienced a taste of web scraping and building our first machine learning project. We were first introduced to BeautifulSoup and Selenium for web scraping, and also the paradigm behind building a machine learning model – Feature Selection/Engineering, Regularization and Cross Validation. From Project 2 onwards, everyone could exercise his creativity and select his own project. Below is the project that I have done:
Predicting the Market Value of FIFA Soccer Players with Regression
Week 4 to 6— Project 3: Classification and Flask App
In Project 3, the Bootcamp soldiered on into Classification algorithms. In this period, we also covered other data science tools such as SQL, AWS and Tableau, however they were more related to the breadth of our data science education, than the project. Finally, we were introduced to Javascript and Flask, where we were tasked to upload our finished classification model onto the Flask App, hosted on Heroku. Below is the project that I have done:
Predicting Satisfaction of Airline Passengers with Classification
Week 7 to 8— Project 4: Clustering, NLP and Recommender Systems
In Project 4, our instructor gave us a whirlwind tour of unsupervised learning (clustering, SVD, PCA), Natural Language Processing and Recommender Systems. Here is where the concepts were covered at great breadth and started to get harder to digest. Because the US Presidential Election was impending, I decided to do a NLP project on Donald Trump:
Strategic Analysis of Trump Rallies with NLP and Time Series
Week 9 to 12— Project 5: Passion Project
Project 5 is also called the Passion Project by the Bootcamp! We were allowed to work on any machine learning projects to our liking. Nonetheless, because deep learning was covered (CNN,RNN) was covered in Week 9 and Week 10, some of the students worked on a project on deep learning. However, as I have not completed a project on Recommender and I wanted to deepen my fundamentals first, I decided to work on a Content-based Recommender. Below is the project I have done:
Pros of the Bootcamp
I would say that the Metis Data Science Bootcamp is truly designed to jumpstart an aspiring data scientist into his or her data science career, if he or she puts in sufficient effort with the right attitude, and also begin the Bootcamp with sufficient background. Apart from providing you with the arsenal of data science knowledge, the Bootcamp is peppered with career preparation workshops, from which I have benefitted greatly.
Moreover, the thinking process, work ethic and attitude that my instructor inspired us transformed our horizons our the 12 weeks, When my instructor said that we would be reinvented at the end of the Bootcamp, he was not kidding. Many of my peers did not think he was either.
Also the Bootcamp requires an entry examination during the application, and thus only applicants of sufficient background will be filtered. This means that you would be interacting with peers who are of the same level, if not more advanced than you. Personally I feel having competent peers facilitated my learning a lot when we share with each other, and at the same time push me to achieve more.
Cons of the Bootcamp
With the being said, the Bootcamp is an intermediate level course. And right at the end of the first week, we are already pressed to complete a full-fledged data science product with only a brief tour of the nuts and bolts of Pandas, Matplotlib and Seaborn. That is where the disadvantage of having the Bootcamp lies. Although it seem to teach a lot, most of the time it only scratches the surface of the topic. Hence, in order to do an excellent and more complex project, one is forced to dig deep into the topic by doing a lot of external self-exploration. For someone less resourceful or less initiated, he or she may end up spending time less productively and settling for less in the project.
With only about half a day spent on fruitful instruction, some may ask if the Bootcamp is truly worth the huge fees that it charges. For me and other peers in the Singapore cohort, it is probably extremely worth it as the Singapore government actually sponsors 2/3 of the Bootcamp fees for Singaporeans. Nonetheless, considering the structure and design of the Bootcamp, I would say that even the full fees are reasonably worth the value if dedicated efforts are spent on self-exploration, projects and landing a job during the Bootcamp.
Advice on Gaining Full Leverage of Bootcamp
If you have read till here, I believe you would have gained a good understanding of the Bootcamp. While the Bootcamp may not work for everyone – some prefer to the take route of conventional instruction by colleges – I have observed some factors that could optimize the learning curve in the Bootcamp, and thus potentially landing one in a job:
- Have sufficient background knowledge before the Bootcamp
This could possibly be the most important factor. Although, the Bootcamp has an entry test to make sure everyone is on the same level, the entry test is still rather basic in nature. I would highly recommend that anyone who wish to enrol in the Bootcamp has at least a working knowledge of the nuts and bolts of Python, as well as the Numpy, Pandas, Matplotlib and Seaborn packages. Some of the peers and alumni have even gone further to learn about machine learning and deep learning in advance! For interested learners, I would also recommend these Udemy courses – Python Bootcamp and Machine Learning Bootcamp – by Jose Portilla. Having enough background knowledge is important, as it allows you to catch up easily with the Bootcamp’s content and dive even further in your projects. Also, apart from programming, I would highly encourage potential students to have good understanding of Statistics (Hypothesis Testing and Regression) and Math (Multivariable Calculus and Linear Algebra).
- Ask Questions during the Bootcamp
Because Data Science concepts can be complex in nature, sometimes it is unlikely that you could understand the first time that the instructor is explaining it. Thus, it is crucial to be bold enough to ask your instructor to explain again or clarify your doubts, before your doubts roll on to next concept. If you instead keep the questions to yourself, very soon you will be lost and will find yourself memorizing rather than understanding what is taught. Also, I cannot overemphasize how a solid understanding will help you in the data science journey in the Bootcamp.
- Build Rapport with Peers and Instructor
While some people prefer to keep to themselves, and in their solitude, come up with amazing projects, I would generally encourage building rapport with peers and the instructor instead. Not only does this make the Bootcamp experience more palatable, ideas may also cross-fertilize during interactions, which give rise to better project outcomes. Another thing to note – tech people are usually very friendly and helpful, and there are a lot that we could learn from them.
- Make Use of the Free Time
I cannot mention this enough. Half a day of instruction at the Bootcamp is certainly not enough to skyrocket your data science knowledge. Thus, in order to maximize your learning experience, it is paramount to revise and continue the journey of self-discovery and reading through Google and other various learning platforms. Therefore, how effective and fruitful the Bootcamp is really depends on how much time and effort you spent spearheading your learning journey.
Last by not least, one could also spend time searching for a job during the Bootcamp. Depending on your preferences, some people might avoid this, as they would like to focus and maximize their learnings during the Bootcamp, which I feel is a worthy choice. However, searching for jobs during the Bootcamp has its pros, as your instructor is likely to offer advice and support should you need them. Moreover, isn’t it also more satisfying to know that you have secured a role upon graduation?
Conclusion
The Metis Data Science Bootcamp has truly been an amazing and transformative experience. Looking back, I was apprehensive about signing up for fear that I could not live up to expectation or that the Bootcamp may not be worth the value. Having taken the plunge and subsequently securing a job as Data Scientist, all I can say is that sometimes taking calculated risk can lead to unexpected and beautiful outcomes. Hence, I would encourage any aspirant who wishes to jump into the data science industry to really consider the Metis Data Science Bootcamp! And to fellow data science enthusiasts and professionals in this exciting field, let us strive on for more exciting possibilities. Godspeed!
P.S. Self-directed learning is an important skill for any Data Scientist, with or without joining the bootcamp. For Data Science aspirants who wish to learn more effectively, I would recommend another article that I have written:
The Complete Guide to Effective Learning in Data Science
Lastly, thank you very much for reading! Follow me on GitHub and also reach me on my LinkedIn or comment here below to discuss!
Support me! – If you are not subscribed to Medium, and like my content, do consider supporting me by joining Medium via my referral link.
