Transitioning into Data Science
Data Scientist is a job that seems to attract people from many different educational backgrounds. I’ve worked with data scientists who have degrees in maths, finance, economics, psychology and music to name a few. Due to the multidisciplinary nature of data science, it makes sense to have this level of variety, but it can also be daunting for newcomers as there’s a lot to learn. Working with a variety of people with different perspectives is also one of the best things about being in this area.
I see data science as the intersection between maths/stats, programming and business knowledge, but it’s extremely rare to have all 3, especially after finishing their studies. A common question I get from people is “how do I move from my current role as X to become a data scientist?”
Digitas has the most diverse data science team I’ve worked in, and that extends to people’s education and career background. 3 members of the team will talk about their backgrounds and give some advice that may help people who are looking to make a career move but need a bit of direction.
Paramdeep Khangura — Senior Data Scientist
Educational background — Accounting/Finance
Career background — Worked in financial analyst roles before moving into pure data science
I see a lot of people from finance, business and economics backgrounds wanting to move into data science and I think it’s a good fit. You understand how a business works and look at data science with a more practical and less academic viewpoint.
Learning a programming language is usually the first hurdle. SQL is a great place to start, especially if your current employer has a database you can query to pull data. Then, learning R or Python should be a priority. While Python is easier to recommend due to its general purpose nature, it’s worth checking what your colleagues in a data-based role are using as it will help when it comes to asking for assistance and working more closely with them (if that’s an option).
As with learning a foreign language, practice is key. Finding datasets, projects, old data science competitions (Kaggle is great for this) and put into practice the new skills you’re learning. Consistency really is the key here and I made an effort to do a little bit every day.
With spreadsheets and Excel being a cornerstone of finance, I found that this gave me many opportunities to use R and Python for automating and improving processes. Data is typically stored in a spreadsheet in a structured format (a data table), and pretty much all the formulas in Excel can be replicated in a programming language. This was perfect for tasks that were done periodically (month end, monthly re-forecast, daily reports). As well as getting to practice what you’ve learnt and speeding up processes, you get the added benefit of there being less chance of human error in the outputs.
I found the approach above really helped me integrate one of the fundamental skills of data science ( programming) into my daily work life, which then helped changed my role and opened up development opportunities in other areas like machine learning.
Divya Rana — Data Scientist
Educational background — Maths
Career background — Started off in Analytics and Marketing Science roles before moving on to Data Science, all within the Media Industry
When I first finished my Maths degree in 2016, honestly, I hadn’t even heard of Data Science. The situation may be slightly better now, but when I was at university the companies who were coming to the careers fairs year in year out were always the same — banks, actuarial firms and so on. None of the routes pre-prescribed for Mathematics graduates particularly appealed to me.
And so, upon moving to London towards the end of Summer, I decided on an entirely different path: I wanted to do advocacy work for a charity, or an interesting think tank or start-up. I wanted a career that was creative, allowed me to innovate and to work for a cause that I believed in.
However, when I began job hunting, because of my background, recruiters began coming to me not with advocacy or charity roles, but with analytical roles, and I started learning more about the field. I began as an analyst, then a Marketing Scientist, and now a Data Scientist, all within the Media and Advertising industry.
I had come to learn that the field of Data Science had all the components I was looking for in a career: it allows you to be innovative, to be creative, to solve problems- and to constantly be learning. I have also been fortunate to combine Data Science with causes that I believe strongly in - such as widening participation in the field, and problems of Bias in Machine learning algorithms (link to video and slides here).
Skill-wise, I started off already knowing R and SQL, but currently I work mostly in Python. I think for young people and graduates, or others just starting in the industry, the best way to learn is to be part of a team which works on the kinds of projects you would like to. I have learned much more since working at Digitas, in a team which regularly employs NLP, and a wide range of machine learning algorithms from XGBoost to neural networks, than I was learning in previous roles.
The team leaders are also great at encouraging the team to constantly keep up-to date with the recent developments and technologies in the field, and employ these in our day-to-day work. As a team, we also regularly look at different Kaggle competitions, working on these separately but coming together for knowledge sharing. For instance, we’ve been currently looking at image classification and segmentation by employing transfer learning on pretrained Mask R-CNN models then finetuning with the iMaterialist dataset on Kaggle, sharing knowledge weekly on our progress, and also entering the Cassava Leaf Disease Classification Kaggle competition.
And so for me, the key to getting into the field is not just learning but doing- employing the techniques you have learned in your day to day work, and if your company doesn’t do the kind of work you are interested in and want to learn, then it is well worth considering a move.
Rupert Coghlan — Data Science Director / Team Lead
Educational Background: Music Technology / IT
Career History: IT consultancy, research, data analysis
There is a definitely a pattern emerging from these posts. A lot of people fall into Data Science from other related domains. I remain another case in point. My early career was in the music industry as a sound engineer. I have always looked for a job that required using interesting technology with a requirement to be creative.
Realising that living in London required a more stable career, I embarked on a career in IT consultancy working for some of the big names in the sector, EDS, HP & Serco. I discovered that working with data gave me a close sense of satisfaction as it did working with music technology. As I became more at ease with working on data related opportunities, I was assigned more and more complex tasks, getting very quickly to the point at which Excel falls over.
It was out of this need to work with bigger datasets that took me on the first part of my data science journey. At university as part of my music technology course, we were taught C++ object orientated programming. Whilst it was a good 9 / 10 years between picking up the programming mantle again, the ‘logic’ and thought processes were still familiar to me and I was able to pick up the R language very quickly and start employing it to tackle challenging data problems.
As I have had the fortune to work with some amazingly talented people, I have been introduced to new methodologies myself meaning that I have had to broaden my technical skills quickly. Day to day, I’m working in Python, R, bash, as well as cloud operations to set up the infrastructure to do our jobs. The broader horizon around automation is also critical and can mean that for simple tasks, we don’t have to turn to data engineers to get a job done.
It’s incredibly important at no matter what level of seniority you are in this domain to ensure that you keep your hand in and remain hands-on. You can lose touch very quickly and veer too far into a strategic domain which can start to divorce you from what’s possible, what’s needed and why. I’m still making sure that I find time every week to learn, practice and hone my skillset and not just because I feel the need, I thoroughly enjoy it!
