17 Aug 21 Data
Written by Daniel Billington, Principal Technology Recruiter - Data
Data Scientists are often seen as the rockstars of technology teams, providing some very niche insights and changes that can help the wider business run more smoothly and efficiently. Current trends tell us that whilst its great to have Data Scientists, they can be underused and often redundant if they cant get hold of the data they need. With this we have seen a huge rise in demand for “Data Engineers”, as one Data Scientist put it to me, “you cant put a roof on a house without having the foundations in place” referring to the Data Scientists being the final piece of the puzzle.
The “best” and most appealable Data Scientists seem to be those who aren’t afraid to get their hands involved with the data journey from start to finish. They go and get their own data and have a true understanding of what is going on under the hood.
This article focuses on Laurence, a now Data Scientist, and his career journey to date. Whilst being a Data Scientist pays the bills, he genuinely has a true passion for what he does and further innovation. Outside of the data world he can be found in his local MMA gym or scooting about on his latest motorcycle. In the past he was manning the doors at some of Lancaster’s most picturesque night spots and more recently he has found himself looking after the newest member to his household.
Although, how do you go from Lancaster door man to Data Scientist?
I studied Physics at University, completing a master’s project in Early Universe Astronomy. I was given the opportunity to work with some of the most exceptional data available, studying a galaxy in the very distant & early universe. I really enjoyed the process of interrogating a data set, performing analysis, modelling, and building strong intuition of the data I was working with. Data Science at the time was emerging into the mainstream, and I knew other people from Astronomy backgrounds made very successful transitions and loved the work, sharing some aspects of approaches to solving problems.
I do, mostly – my degree experience (and especially my masters project) has given me confidence in solving difficult problems and remaining pragmatic when doing so. It has also taught me how to learn quickly and effectively and required me to get up to speed quickly with Machine Learning theory to a good standard, without getting bogged down in unnecessary detail.
On the other hand, (perhaps unsurprisingly) it did nothing to prepare me for understanding the way businesses work, and how to effectively run a project in industry. Every Data Science job I have had has taught me something new on that front. Despite it being a python project, it also didn’t help on the coding front – looking back my code was truly horrendous!
I don’t believe a degree is essential. Data Science is more about how you think, understanding principles, building intuition, asking the right questions to understand the problem, effectively communicating, and having a well-stocked technical toolbox.
I have worked in three different companies since graduating, having worked at a tech giant, a small fintech and a huge logistics company. All of them have provided me with a completely unique set of experiences and have broadened my perspective on being a good Data Scientist. The tech company provided me with the greatest learning opportunities in terms of applying advanced Machine Learning. Pace of delivery wasn’t particularly fast however. The small fintech provided me with the highest pace of delivery of consistently commercially successful projects. The company invested heavily in Data Engineering and will pay dividends for Data Science innovation. The large logistics company provided me with great opportunities to work on collaboration, expectation management and stakeholder communications, all in a company where Data Science was very new.
I think that I have been lucky to recognise what skills and knowledge are and aren’t important for (most) Data Scientists to possess. I don’t spend hours trying to memorize mathematical equations or read every Machine Learning paper claiming to be the state of the art (been there done that – trust me, it doesn’t help!). To successfully apply Machine Learning you must understand the best practices which underpin the model training and evaluation process. These, thankfully, are much easier to learn, remember and apply and will consistently add value.
If your team fully own the product you are creating, building a model is just the beginning of a project. Data Analysis combined with an understanding of the business are required to figure out the best ways to use a model. It also needs to be deployed and integrated with systems so that it can operate smoothly. This is where ‘full stack’ experience is developed e.g. BI software to build model reports & monitor live systems or using Cloud Technology and DevOps tools to robustly productionize models.
In terms of pushing myself, I think it is good to keep changing up what you are doing if the time is right. Personally, I see no value in staying in a particular role for an arbitrary amount of time for its own sake, if there are better opportunities out there. Especially if you have delivered a lot & learned a lot. I feel that I have reaped the rewards of moving to different companies and exploring new ways of working and applying machine learning in a new domain, with all the challenges and hurdles which are part of that.
Currently I don’t have any major goals that stand out. The future of Data Science and Machine Learning is an exciting one, and day by day the less glamorous aspects of the job are getting easier with new Cloud technology, leaving more time for solving problems. For now, I am happy to soak up as much experience as possible in applying Machine Learning to business problems, and keep gradually adding to my toolbox with the latest tech. With each job that I have had, I have picked up on what ingredients lead to a smoothly operating project, so perhaps branching out into managerial aspects in future is something I will consider.