If you’ve read our other blog posts, you might have realised innovation is at the core of everything we do – and the Data Science team is no exception.
The recent growth and adoption of data into various industries has brought an onslaught of MOOCs (Google it if you don't know what it means 😉) and specialised programs churning out an abundance of bright-eyed candidates looking to use machine learning to solve real-world problems.
The issue here is that this isn’t the bread and butter of what Data Science is and you lose sight of providing innovative solutions and blindly rely on algorithms.
So what actually is Data Science? 🤷🏻♀️
Data Science is a discipline combining technical prowess in mathematics and computer science with a deep understanding of the problem space to craft insights and solutions.
As practitioners, we then have a toolkit of employable methods such as statistical analysis, network analysis and machine learning. At Revolut our only goal is to maximize impact and minimize our time to deliver; in many cases this can be achieved with strong statistical analysis on the data. But once we’ve hit the boundaries of our existing solutions, we go and explore the benefits of machine learning.
At Revolut, our Data Scientists and Data Engineers are not back-office geeks churning away at dashboards, ensemble models or data pipelines without ever being able to realise the quantifiable impact of the 'n' months of work. Rather, they are deployed on the front-lines to work directly with or as Product Owners to deliver solutions. Successful members of the team at Revolut gauge complexity and build solutions that are:
✅ High Impact
Tell me more about the team 🤔
Our data team is quite young – in fact we just hit our 1 year anniversary!🥇 After joining as its first official Data Analyst, I noticed a lot of inherent biases in decision making processes and realised that we needed to be more data driven. Due to our rapid approach to delivery it soon became apparent that we needed to build a team – which now consists of Data Scientists and Data Engineers!
Revolut deploys an embedded team structure across the company, so the data team members are allocated across several different functions with the option of rotation. Currently members are deployed across:
✅ Platform – Build kickass infrastructure to automate and accelerate solutions
✅ Fraud & Anti-Money Laundering – Fight the bad guys
✅ Finance – Mitigate loss & optimise processes
✅ Growth & Engagement – Improve quantity & quality of users
✅ Credit - Building the next generation of credit models
✅ Talent Acquisition & HR – Intelligently hire & improve quality of work
In the near future we hope to add:
🔜 Customer Experience - Intelligent agent allocation and chatbot optimisation
🔜 Premium - Improve and implement new features to increase revenue
Despite this system, our team is extremely collaborative and we consistently strive to avoid doubling up on tasks and share all the cool 'sh🤫t we got done'.
The Data Scientist 👩🏼🔬
Revolut’s Data Scientists bare a lot more responsibility than one at a larger company. Generally, there are several steps involved with providing viable solutions, which include but are not limited to:
- Forming a deep understanding of the problem space
- Identifying viable data sources and conducting rigorous exploration of data points
- Architecting a framework (model or process) to solve the problem
- Feature engineering (remember this does not only have to be for ML!)
- Building data pipelines
- Developing/coding the model or pieces of the process
- Productionizing and deploying
There are unicorns amongst Data Scientists who are able to do all of the above, and although those are rare, we do have some in our team – so at Revolut we offer opportunities for Data Scientists to learn from Data Engineers to fill in the missing gaps, should they desire.
The Data Engineer 👨🏻💻
The challenge with data engineering at Revolut is keeping up with the exponential growth of the business. We are crazily data driven and it is surprising to realise what departments a data engineer may help (e.g. setting up data reporting for the legal team).
To fuel our growth, our projects move at extreme acceleration. In order to ensure no one is blocked, what takes weeks elsewhere should take hours here and what takes hours elsewhere should take minutes here. Besides the speed, we have to assume that every data model we develop will quickly expand in volume, complexity and variety - this has to be factored into solutions at the very beginning.
We simply cannot afford conventional data engineering approaches here, those will quickly become unmanageable and fail the company. Data engineers here don’t write data pipelines, transformation processes and models - there are too many of them to maintain.
We must be creative here. We take a step back and go one level up. Instead of writing pipelines, we write meta level code which generates data pipelines, transformation processes and models. Also, since we are growing so fast, we try not to rely on any specific storage or transfer technology - we can switch quickly from an analytical database to another just by creating another implementation of an abstract class.
To sum up, our data engineers adopt the following principles:
✅ Don’t repeat yourself (DRY)
✅ Configuration over code (and convention over configuration, of course)
✅ Declarative design
Our Vision 🚀
This year, Revolut is aiming to build out one of the strongest data teams on the market and to innovate future-proof solutions. The buzz and hype around data is a clear indication that traditional roles & titles will change in time, meaning we should be ahead of this curve.
We have already armed a few teams with the power of data science and plan to integrate it into many more product teams as we develop.
Join Revolut for Free
Manage your everyday spending with powerful budgeting and analytics, transfer money abroad, spend easily in the local currency, and so much more. Join 18M+ already using Revolut.