My Journey From Account Handler To Data Cruncher (Via Techno DJ)

During my time in the client services team I worked on a number of different market research projects. This gave me a taste of how data could be used to draw meaningful insights. I wanted to explore the world of data science further but I was initially intimidated by online research suggesting I would need a PhD and 5 years of experience just to get my foot in the door.

The decision to change my career path didn’t happen overnight. It was something I’d been mulling around in the back of my mind for a while. As the position didn’t yet exist within the London office, I knew I’d have to create my own path.

At first, I didn’t know where to begin my research. I didn’t know a thing about coding and I hadn’t touched statistics since completing my Neuroscience degree. The feeling of nausea from all the uncertainty and unfamiliarity didn’t budge, but I went for it and have no regrets. To tell you about my experience, I’ll share some of the key steps on my journey.

1. Selecting a tool/language

The first step was deciding which tool/language I wanted to learn. I’d read that Python was a versatile, high-level programming language with multiple libraries, meaning that it could be applied to a number of different roles – this seemed like a good place to start.

2. Choosing a course

I started to do some of the free online Python tutorials a few evenings a week. These were great at covering the mathematical foundation and then applying it to data science-related Python programming. What I started to realise was that, although I really enjoyed solving daily code academy challenges, I wasn’t really grasping the theory behind what I was doing.

3. Finding a network

At this point, I’d reached a bit of a hurdle. I wanted to take my studying to the next level but reading about machine learning theories was far too overwhelming. I also knew that I needed, not only good instruction, but a good network of other people in my position. I started by searching ‘Python’ in my LinkedIn to see if any of my peers had this listed as a skill and would be willing to tell me more about their journey, but I didn’t have much luck. My next step was looking at Meetup, where I found an ‘introduction to data science’ event held by General Assembly. At this event I was lucky enough to meet a group of people in my position and get advice from others who had been there the year before. The speakers at the event were so helpful at providing resources and guidance that, although I ended up enrolling onto a boot camp by a different organisation, I am still in touch with a few of them today.

4. Taking a leap of faith

After deciding to take a 3 month sabbatical to enrol in a boot camp with the support of Saatchi & Saatchi Wellness, the next step was completing the 120 hours of pre-work. This spanned linear algebra, statistics, python and git. Finding the time to complete this outside of the inevitably long working hours of agency life was a huge challenge and there were more than a few times where I wanted to give up. If I was struggling with the pre-work how would I keep up with the intensive, fast-paced environment of a boot camp with a curriculum covering technical topics I’d never even heard of? With a lot of planning, prioritisation, and knowing when to take a much needed break, I managed to complete all the pre-work before the course started and made it to boot camp.

Looking back, I am so glad I decided to take the leap and apply. Even though I admire all those who taught themselves solely through online resources, the expansive curriculum and collaborative environment that I experienced at boot camp was invaluable. I was amazed by how quickly I gained the confidence to write my own scripts and learn technical concepts that had evaded me for so long.

Over the 12 weeks, I developed 5 machine learning projects. My favourite project to work on was our final ‘Passion Project’. I created an AI DJ able to generate an original techno track based on an existing playlist. This saves DJs a lot of time and effort searching for new tracks to fit their current set list.

To do this I developed an LSTM (long short term memory) Neural Network using the Keras library. LSTMs consist of gates which allow it to learn which information is important to either keep or forget. This makes it useful for working with sequential information, such as music. I trained the model on Techno MIDI files and it learnt to predict the next note in a sequence, thereby generating new music.

I’m now so excited to be back in the London office as a data scientist. Here, as part of the Integrated Strategy and Engagement team, I will be providing clients with data driven recommendations to help build, launch and grow their brands. The journey was an incredibly eye-opening and liberating experience, and I recommend anyone considering retraining to take the plunge and go for it.

What do you think?

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: