How to transcend/start in Data Science/Machine Learning
What worked for me in this amazing journey of discovering Data Science and making a career out of it. If that’s what you are looking for, I’m sure it would help you!
To begin with, let me introduce myself. I currently work as Machine Learning Engineer at Mckinsey and Company and have 4 years of experience. I completed my graduation in electronics engineering (2017) and currently pursuing Masters in Artificial Intelligence from IIT-Jodhpur.
I’ve been working in the corporate with organizations solving business problems that have the potential to use advanced analytics practices leveraging machine learning and deep learning technologies. I have worked on cool engagements where I worked on Fraud investigation!
For anyone who wants to break into the Data Science domain, the journey hard. The reason being, firms require a lot less Data Scientists than Software Engineers. And since most of these roles exist for revenue optimizations, sales, forecasting which eventually translate into big business decisions, the stakes are high. Hence the requirements are steep. Being familiar with everything is not sufficient, you need to be good at it. And to be that good, it should be not that hard lest sufficient time has been invested into it. Here I will break down what you can do if you want to break into the Machine Learning paradigm given the kind of experience one has and what to do after that.
Step 1. Who are you?
Identify yourself. Do you know to program? If yes → do you know Python? (I am not an advocate of R because of its own limitations) → Yes? Let’s do it then! Else if you are not familiar with coding or Python programming language I would recommend these resources to get you started :
a. Python Language from NPTEL (Chennai Mathematical Institute)
b. Udemy: Python for Machine Learning
( I have provided these resources and any one of them should suffice. Once you get the hang of the language better move ahead. Completing the course should not be the objective. Treat this as a stepping stone.)
Step 2. Let the machines learn!
To study Machine Learning, the best resource is Andrew NG’s course on Machine Learning. If you want to go for a certification, you can sign that upon Coursera (which I recommend) else you can view all his lectures on YouTube. The course has assignments that are in MATLAB which can surely be avoided, but the same assignments are available in various GitHub repositories which are ported to Python. This will be a long journey definitely. I hope you’ll stick around until you have all the intuitions you need to go to the next step. try to get all your questions answered from the community forums or even stack overflow.
If you are familiar with some of the basic nuances of machine learning and want a code-first approach towards learning, you can go to Machine Learning A-Z™: Hands-On Python & R In Data Science in Udemy and check it out. It’s also not a short course but will give you a lot of coding practice along the way.
Step 3. Practice! Practice! Practice!
Now you know the algorithms, now you know also know the intuitions. What now? Let’s practice.
Kaggle is the best platform for start learning problem solving using Machine Learning algorithms. You can spend days just watching other people’s work and wonder how the same problem is being approached in so many ways!
To navigate Kaggle, start with replicating other peoples’ work. Once you are comfortable with the flow, then try it on your own (the same problem) discuss the in the forums and see if what you think really makes sense and what does the output of your model really mean.
And later start competing! You really don’t want to win the competition but I like your attitude!
Step 4: Level Up!
Now that you are familiar with Machine Learning and how Data Science uses ML concepts for predictive analytics. Let’s go for the much-hyped and considerably complicated Deep Learning. Again I’ll recommend Andrew Ng’s Deep Learning Specialization at Coursera. I would strongly suggest going for the complete certification because the assignments in the curriculum are excellent.
The only reason why I separated ML and DL from each other because I want the user to practice first the (now easier) ML concepts so that when he/she gets started with Deep Learning, you don’t have to worry about coding but more about the problem and the model.
After all this, you should’ve started to apply for roles/internships in companies that interest you. Preparing for these will be a whole separate section and I would probably discuss this in a separate post.
Bonus: Follow up!
By now you should have an idea about what interests you in Data Science and become a specialist in at least any one of them. If you like text, take a course on text analytics or computer vision if that excites you. And if the stars align well, you’ll get the opportunity to contribute and work at various levels in the industry. The next steps would be to study cloud services and productionalizing ML.
I understand this could not probably be the best resource for anyone to get success, but these are all the things I followed during my journey. I was also working on live projects so I did that using my own sweet time. Although I must admit I’ve read a great more deal of it than what I’ve mentioned above, mostly it was part of the work I was doing and as an introduction, these resources should suffice to intrigue and sustain your interest. In hindsight, applying what I’ve learnt has always made me a better Data Scientist than just knowing. But anyone starting from scratch would take anywhere between 6 to 12 months.
Don’t get discouraged! It’ll be all worth the effort! I’ll still say I’m learning and never an expert. Machine Learning has its own ways to make us humble 😛.
Below is my link to LinkedIn where you can connect with me and we can discuss any problem, suggestion or discussion around all things data.
LinkedIn : https://www.linkedin.com/in/shantam-saxena-95b068103
