In this career guide, you can learn more about what a job in Quantitative Finance and Data Science (with focus on finance) entails. These industries are perfect for those who prefer professions that require a more technical approach and have a strong understanding of programming, maths and statistics. Below you will also find some useful resources and advice from UCL alumni who have worked in these industries.
Table of Contents
Jobs and Salaries in Quantitative Finance (QF)
Quantitative Analysts, or Quants, usually work at investment and retail banks, asset management firms and hedge funds. There are several types of quant jobs.
- Front-office Quants provide traders with pricing or trading tools.
- Risk Management Quants perform risk analysis on assets and markets. Their goal is to find the perfect balance between the appropriate risk that will maximise the returns.
- Algorithmic Trading Quants use highly complex mathematical models that enable to automatically buy or sell a certain asset at a predetermined price and time.
- Quant Strategists, or Strats, carry out research and create new trading strategies that can be implemented by the firm.
- Quantitative Developers, or Quantitative Software Engineers, are highly proficient programmers who implement models and strategies created by the quants/strats.
Average salary in London (per year)
Jobs and Salaries in Data Science (DS)
While QF is a more mature industry, DS has been booming over the past few years. Sometimes people use “data scientist” as an umbrella term that also includes data analytics and engineering. Briefly, data engineers clean the data and store in an accessible way in a local database or on a cloud platform. Data scientists model, interpret and visualise that data. Data analysts do similar tasks to data scientists; however, their line of work usually involves less programming and more data interpretation.
Nowadays, there are also machine learning (ML) engineers who work closely with data scientists. As the name suggests, they use various ML techniques to understand the data and unveil the trends it holds.
Average salary in London (per year)
Source: Indeed. Note: these are general salaries that are not finance specific.
Skills to have
There is a huge overlap between the technical skills in QF and DS. For example, a data scientist working at a bank might be doing tasks very similar to those of a quant.
Here you can find some of the topics that you should be familiar with when applying for a data scientist or a quant role.
- Linear algebra is the foundation you must have. Brush up your knowledge in vector and matrix manipulations, eigenvalues and eigenvectors, etc. The same applies for calculus: integration, derivatives and differential equations are very important for understanding of the main concepts in DS and QF.
- Probability and statistics are another must-haves in these industries. If you feel that you need to do some revision or learning on these topics, don’t worry – there is plenty of material online. For example, MIT offers a range of free online courses: Introduction to Probability and Statistics, Statistics for Applications, Probabilistic Systems Analysis and Applied Probability to name a few.
SQL (can also be pronounced as “sequel”), a data-base programming language, is another a frequent requirement for DS roles. To learn more about SQL structures and common functions, you can check out mode.com.
In addition, make sure that you are familiar with different data structures and how they are implemented. You also need to understand code complexity (the Big O notation); this website provides an overview of different space and time complexities (check out their comment section for additional resources).
Some roles might also require familiarity with legacy tools, such as Excel and VBA
- You can practice interview-like coding questions on LeetCode, HackerRank, and participate Kaggle challenges.
- ML has boomed over the past few years and is now a common tool in DS and QF. Again, there is a wide range of different courses and other resources where you can learn ML. An example is Google’s ML crash course or Stanford’s ML course on Coursera. You can also watch educational videos about ML; for instance, the YouTube channel StatQuest has great videos on a range of ML topics. You might also want to look at this book on deep learning, which, in addition, covers the main concepts in linear algebra, probability and statistics.
- For some QF specific resources, look into these MIT courses: Analytics of Finance, Topics in Mathematics with Applications in Finance and Finance Theory. These will help you get a solid understanding needed to successfully pass a QF interview.
- Don’t forget GitHub. Having a public repository on GitHub is a great way to showcase you programming skills. This tutorial can help you get started.
Interview with a quant
Aakash Karia graduated from UCL with a BSc in Physics in 2020. During his university studies, he interned at RDP Newmans, an accountancy and taxation firm, and tutored GCSE students.
– You have done an internship in accounting and taxation, while your degree subject was physics. How did you end up doing quant finance?
– I believe the main reason why I chose quant finance was the opportunity to implement the skills I had acquired during my studies. A degree in physics doesn’t only provide you with technical knowledge like maths but also gives you a range of transferable skills, such as the ability to solve abstract problems. You learn how to simplify these complex problems to explain them to a diverse audience.
I also enjoy working in a collaborative environment where you can interact with your team and develop your leadership skills. Unfortunately, I didn’t feel that physics offered that experience because during my degree, everyone was very enclosed in their own studies, in their own book, and that put me off a bit.
In addition, I’ve always had an interest in the financial markets and I’ve done some equities trading on my own. So, I was looking for a career that would combine finance with my technical and abstract thinking skills, and where you also get to interact and explain what you’ve done to an audience. Quant finance combines all of that.
– You’ve mentioned that you’ve done equity trading. Did you do it algorithmically or manually?
– I did it manually, but I was using technical indicators in my trading, for instance moving averages, various stochastic indicators and oscillators. And if you actually dive deep into them and learn the equations behind them, they turn out to be very similar to what you learn in physics.
– In your current role, you work with risk modelling. What does it entail?
– We basically use decision science to create risk models. These models are then used by other business teams for their individual portfolios. In my current role, I mainly work with commercial and retail banking products, including lending to SMEs. The models we use are called credit risk scorecards. In simple terms, our task is to estimate the probability that a company or an individual will default on their loan. Our goal is to find the perfect balance which will minimise the risks for the bank, while maximising the upside on the returns. This is essentially a support role, so it’s a part of the middle office. Some people believe that the risk team always tries to limit down the front office because of potential risks. But what we really do is find the optimal risk that will increase the bank’s returns through fees, interest rates, etc.
– What hard skills do you need to succeed in a quant finance role and how did you acquire them?
– In my role, you need a solid understanding of probability and statistics as well as programming skills. In my studies, I used Python, but at work, we mostly use SAS and Excel, although there is a transition towards Python at the moment. My degree in Physics gave me the foundation of core concepts, which made it possible to further build new knowledge on top of it. For instance, I didn’t know any SAS when I first started working, but it is easy to learn if you know how to use Python because the main difference lies in the syntax.
– Do you have any advice for current and future applicants?
– For your CV, have a specific section for your coding skills and list any projects where you’ve demonstrated your data analysis skills, statistical and mathematical concepts, and ML. At the interview, they will give you chance to speak about these projects to show your skillset. I’ve also learned a lot from attending events organised by economics and finance related societies at UCL, with the UCL Investment Soc being one of them.
– Did you have any public repository on GitHub when you were applying? Also, what was your prior knowledge in ML?
– I didn’t have a public repository, but I should have had one. Instead, I primarily had projects from university and an internship that I could talk about at the interview. Regarding ML, I was familiar with clustering and logistic regression methods. Currently, there is a shift within the industry towards gradient boosting, so be aware of that.
– Were you offered any courses or training when you started working?
– I was offered SAS training organised by a firm called Amadeus. And in the future, there will be an opportunity to do more advanced training on SAS.
– If you were to name one thing that you love and one thing that you hate about your current role, what would you say?
– One thing I love is that my work tasks are very diverse and every problem is unique. It makes my job interesting and allows me to use and improve my analytical and problem-solving skills. It also gives an opportunity to be creative while finding the best possible solution to a new problem.
One thing that I dislike concerns the regimented process that you encounter at a big corporation. You are required to follow specific procedures, with some of them set in stone 5-10 years ago. At a smaller company, you’d have more freedom in how you approach a certain problem.
Interview with a data scientist
Umais Zahid completed his BSc in Physics at UCL in 2019 to start his career as a data analyst at Deliveroo and later as a data scientist at the credit management firm Cabot. In 2020, he returned to UCL to complete a MSc in Computational Statistics and Machine Learning which he very recently finished.
– Why did data science appeal to you?
– You could say that physics was my first love, but in addition to that, I’ve always been interested in neuroscience. I got progressively more into neuroscience during my undergraduate at UCL, and this is how I got into AI as they are a lot more adjacent than you’d expect. My final goal was to end up in neuroscience or AI research, and at first, I planned to do a related masters directly after my undergraduate. However, I realised that I couldn’t afford it, and I had to get a job that ideally would be related to my goal. Data science was a step in that direction since many of the tasks you do there are similar to the tasks in computational neuroscience and AI. I also knew I had the skillset for it: I’ve done quite a few Python-focused modules at UCL where I got exposure to Jupiter Notebooks and useful packages like scikit-learn and pandas, for example. This skill match prompted me to apply for data science roles.
– Was it difficult to land the first job offer?
– My first job as a data analyst at Deliveroo was a two-month contract, and I was referred by a friend. The second job as a data scientist at Cabot was harder to get, and the application process took about 4-6 weeks. It felt like a very protracted period of time back then, but now I know that it is actually quite normal. I’ve gone to the final stages for some of the data science roles I had applied for, but usually, they wanted someone more senior. At the time, it wasn’t very common to enter data science directly after an undergraduate; maybe things have changed now. But I knew I had the required skill set; I just couldn’t convince the employers. To resolve this, I started working on some projects in my free time, and that’s how I eventually got on offer.
– What were the projects?
– The first project involved creating a credit risk model. It was a relatively simple project that was pretty self-contained inside a single Jupiter Notebook. To expand on this, I also created a bunch of different models for different data sets. Thanks to this project a recruiter contacted me about the role at Cabot. The second project was quite significant for my interviews as well: it was a web scraper for the jobs on Indeed.com, for which I created a web app with a front-end interface. This project was more or less straightforward and took from a couple of days to a week to complete, but it made a big difference and showcased my skills. The third project I did about a year prior; it was a trading project which is also available on my GitHub. I basically connected a consumer trading platform (MetaTrader) written in a C++ type of language with a Python interface. I was also planning to do some machine learning with it but got distracted by my course work.
– How did you come with the ideas for the projects?
– For the credit risk model, I was looking for topics where DS and statistical modelling have been commonplace for a long time, with credit risk being an obvious choice. In credit risk, models basically predict the probability of default (or alternatively, liquidation). The job scraper was born out of necessity: when I was looking for a job, there were many roles that I wanted to filter out, headhunting jobs, for example. So I created a script that would exclude certain keywords, like headhunting. The MetaTrader connector project was more of an experiment; I wanted to see if it’d be possible to obtain very high-resolution financial data to do some analysis with it.
– Did you use any ML in those projects?
– Yes, in the MetaTrader connector project. I had very little experience with PyTorch because I took an optional module in ML at the Computer Science Department in my last year (which I highly recommend!). But actually, all of the ML in my role at Cabot was done in scikit-learn (linear regression, random forest, decision trees, etc.). I’d say about 90% of industry roles in DS will not require a lot of ML. Especially in finance, where you have a large number of stakeholders and a lot of money on the line, the models must be transparent and interpretable. For now, many ML models are, on the contrary, fairly black-boxed.
– If you were to name one thing that you loved and one thing that you hated about your previous role as a data scientist, what would you say?
– I loved that your day-to-day tasks are quite varied and unexpected. You often deal with “dirty” data, which requires a lot of data cleaning, but you also deep dive into the problem to uncover insights. It involves a lot of statistical exploration: dissecting the data in various ways until you understand what’s happening, whether there are clusters of populations that behave in a certain way, which you need to model significantly differently and segment them out. Generally, I like DS because it offers a mix of statistical analysis and software engineering.
One thing that I disliked was related to the fact that large financial institutions and banks are rather rigid and bureaucratic in their approach. There is a fair amount of resistance to adopt new methods, even if they are more efficient. I think it’s less of a problem in FinTechs and start-ups. That’s why I’m now trying to shift towards more tech-related roles instead.