People afraid what they do not understand, but worse is we have fewer people attempt to learn them in the first place.
This article is about my first two weeks experience of doing an internship as a Data Scientist at Data Wow where I start off from scratch. I decided it will help out others college student who is still struggling with what they should do with their internship as well as others who might find this useful.
I reached out to a startup company called Data Wow. Fortunately, the company took me in despite me saying I have zero experience in this.
Since it is a startup and we got readers from all around the world. First, I would like to give you a background on this company where I do my internship, or Data Wow is an Intelligent Outsourcing Platform and Machine Learning. It is located in Thailand, Bangkok, near Phrom Phong BTS station. It only takes you a good 10 minutes walk to the building.
To be honest, what I find most important is the working environment, and Data Wow offers a solution in every aspect. They have a flat-work structure, everyone is passionate, and more importantly, they work as a team.
Ok, but you went on an interview with a thing you have no experience in? Absurd!
Yes, It was crazy. It took me good couple of weeks to do research on, watch some videos of online courses. To me, regardless of fields that you want to pursue in Science. You will need passion, a right mindset, and a good logic to work with. Time spent for people to achieve their dreams may varies, but as long as you have those three, you will become successful eventually.
You can do it, now it is time to move out of the comfort zone!
So how's the first day?
Nervous, put that in all caps too. Remember your first day at university? This is oddly similar, It almost felt like living in a completely different world.
I was anxious, thinking that this might be the worst decision I have ever made. I should have gone back to my comfort zone, to what I always do.
Surprisingly, It is not that bad, at all. Take that sentence back and put a negate in front of it.
That day started off with a typical orientation, I was introduced to the teams, the company structure and such. Later on, I have a small discussion with them so they know where I am at in term of background knowledge about Machine Learning.
That went awkwardly. Generalisation as a theory? Sure. Explain it to a 10 years old kid? That’s not what I expected to happen on my first day. It made me understand that a good measurement of your knowledge is not how many books you have read, how much you have learned. It is whether you can give out a good explanation to someone who literally has no clue about it or not.
To help me get started, a daily lecture is assigned to me (and a homework). We will do a summary together as they do want to see if I truly understand the lecture. Any questions? Just ask. No need to be afraid.
The next day, and on ward.
I know that they do Daily Scrum, but to have me participate in it as well is a surprise. Daily Scrum is when the team holds a meeting, and they will discuss what they have done yesterday, what are the problems or success they found, and what are the goals today. I was nervous at first since I don’t have much to talk about. At the same time, it is such a great opportunity to have because you get to learn in a real working environment, how the team tackles the problem. Every day is filled with amazing thoughts and ideas that sometimes I wish I can jot down everything they say.
After the meeting, guessed what. We go back to our work.
I thought that this is going to be extremely painful. Seriously? Daily lecture? Turns out it is not when you have someone doing it with you. A senior of my team will help me go through it, and often time we end up laughing at how ridiculous things are in some theory. I did not mean that in a negative way, but It is similar when you do a group study, then one of the people figures thing out and be able to explain it in a short, quick minute from an hour lecture video. Although I have to admit that I do struggle a bit as there is a lot to take in, and we are doing it daily. Sometimes I need to rewatch or reread three to four times to really understand what is going on.
It sounds boring. I know but, not like you can go and take down the boss while you’re under level. A first few stepping stones are needed before the actual fun begin. No, that’s not a sarcasm. It is actually fun.
It took almost 2 weeks to finish lectures and some coding homework up to where I can begin to help them with project. This is where the fun started.
Now, tell me about the fun
Pair Reading paper.
Sorry, but I am serious. To be more precise, I finally get a chance to explore what’s behind the door without getting block by Oh my goodness I don’t understand a single word! Someone save me!.
Now that I am more familiar with machine learning. A senior in the team discuss to me about the problem they are facing in one of the projects, and whether this approach should be useful to solve the problem. I was told on the concept only, then I ask them further whether there are papers or videos related that they can recommend it to me.
"I got a bunch of it but I thought it will be too scary”
It is about Generative adversarial network or GAN in short. GAN is a relatively new method that was introduced by Ian Goodfellow and other researchers at the University of Montreal in 2014. To keep it simple, it is a game with two players going against each other, and both are neural network. For those who are familiar with game theory, Minimax Game is used in GAN. The two neural network will try to outperform each other until they reach the equilibrium.
I will try to summarise the concept in a few paragraphs, and make it basic for those who do not know about machine learning.
Let’s one network be D, the good guy, and the other network be G, the bad guy.
The good guy will be given a dataset, which is composed of the actual dataset mix with data from the bad guy where he purposely puts noise in it to fool D.
Goal of D is to distinguish between the real data and the fake data from G, while G’s goal is to make D fail to tell the differences between the real data and the fake data he made.
You can see where this is going, soon the two will collapse to the equilibrium where D fails miserably to do its job as G made the perfect copy of real data.
I’ll leave the link to the paper if anyone is interested. It strongly recommends it as it is a really good read, especially if you love deep learning.
We have a lot of fun going through the paper. We discuss almost every bit of the paper, his theory, his proof, and lemma (Because we are one’s curious Data Scientists!). Googling up terms or unfamiliar phases. Questioning why or how the author chose to do this approach and why it works. Then we talk about how we can adapt this algorithm to solve our problem. How GAN may actually help us or did we just run into a dead end again.
Questions and doubts go back and forth as we try to crack this black box of problem. It was the most entertaining theoretical talk I have ever had with someone.
Remember the three concepts I told you before? Passion, mindset and the logic. I think I found the first one.
This is my first time writing article, I do hope you find it interesting and thanks for giving this article a chance!
If anyone is interested in working with us, not just Data Scientist but also as a Software Engineer, please contact firstname.lastname@example.org. Links:
Data Wow Company : https://datawow.io
Drop us a line and we will get back to you