Measuring risk
An interview with D-PHYS alumnus Dr Szymon Hennel, formerly risk model developer at UBS.
What considerations and decisions led you to your job in the financial sector?
I joined UBS straight after my doctoral degree. I was part of a team that develops models for computing the financial losses expected under different economic crisis scenarios, a process known as stress testing. The goal is to ensure that banks conduct their business in a way that allows them to continue providing their services reliably even at times of economic crisis.
When I started my physics degree, I thought I would become a researcher studying fundamental questions around the nature of the universe. At the same time, I had always been drawn to information technology: my hobbies as a teenager included programming and tweaking Linux setups. Perhaps predictably, the rise of quantum information processing as a major field of research in physics caught my attention and shifted my interest towards more applied topics in physics. The aim of my doctoral research was to verify theoretical predictions implying that a particular type of semiconductor exposed to high magnetic fields could be used for the first practical demonstration of a fault-tolerant topological qubit. Even though my project was motivated by a big goal that would have a huge technological impact, my daily work revolved around specialised topics of interest to a few experts worldwide. This is not unusual in scientific research and didn't come as a surprise, but by the end of my doctorate I observed that keeping my focus on one topic of purely academic interest felt very different from the learning process throughout my bachelor and master: I didn't enjoy the narrow thematic focus as much.
I could see two options for my next professional step. One would be to pursue postdoctoral research, switching to what is now called quantum engineering and working in larger research communities closer to industry. The other option would be to find a job in the private sector. As I felt more affinity for programming than for building hardware, I figured that an industry job focused on data analysis would be a good fit. Based on the advice I received, I also concluded that gaining experience in the private sector would open more opportunities and grant greater flexibility in the longer term.
“Even if you have an academic career in mind, it's helpful to think of what else interests you and reach out to people early on to create networks.”Szymon Hennel, D-PHYS alumnus
I had become interested in finance by the end of my bachelor's studies, initially because I wanted to be able to better understand some news and opinion pieces. I read up on different topics every now and then and developed a habit of reading the financial press. I also took Professor Didier Sornette’s lecture on financial market risks, which gave me the impression that a job in financial risk management would be an interesting combination of technical challenges, social relevance and intellectual appeal. This field-specific interest and acquired knowledge put me in a comparatively good position to find a job in the financial sector after my PhD. Risk management divisions across the finance industry were strongly growing at the time (in response to regulatory expectations), so most open positions for quantitative work in the industry were in risk modeling. Among the offers I received, I chose to join the UBS firmwide risk modelling team because I really liked the people and the atmosphere during the interviews. I also liked how the job blended financial and economic topics.
I should add that I've recently left my job at UBS to focus on developing my own idea to turn it into a business. This decision process felt very different from choosing a career path after my doctoral studies. I can judge the trade-offs and consequences of professional decisions in a much more informed manner, something that wouldn't have been possible through the advice of others right after my PhD. In fact, I know D-PHYS alumni who turned their research work into startups. Towards the end of my doctoral degree, I joined a team at the business concept stage of Innosuisse's Startup Campus training to have a closer look at the entrepreneurship environment. I didn't have a business idea at the time, but I kept in mind what I learned about the startup ecosystem.
Based on my experience, I'd advise physics students to pursue industry internships before they finish their degrees. And if you start a doctorate and have an academic career in mind, it's still helpful to think of what else interests you and reach out to people early on to create networks and alternative options.
What were your main responsibilities and tasks at UBS?
Let me describe the typical work of a risk model developer in a large bank, most likely in a team who develops models used to determine the amount of capital a bank is required to keep on hand, as opposed to lending it to clients. In this role, your main responsibility is to ensure that a specific model is available for use. In practical terms, you make sure that there's a piece of software up and running on a dedicated IT platform that can be used to calculate certain financial losses (for example, how much money may be lost within the next quarter because clients are unable to repay their loans? How much will the bank lose due to transaction processing errors? And so on). Roughly half of this work is programming; the other half consists in speaking with different stakeholders to get feedback on the model or request necessary approvals.
Depending on the company, performing the computations and filling out reports that are used by the teams who manage the bank's capital or are sent to the regulators may also fall under your responsibility. Sometimes there are separate, dedicated teams taking care of such tasks. Interestingly, banks are required by regulations to have separate teams tasked with validating models. These colleagues verify your work, looking at the mathematical aspects but also considering administrative formalities. Banks are subject to detailed regulations that shape your daily work. There are clear rules for when the validation teams need to be involved, for instance, and ideally this leaves no room for ambiguity.
What does an average week as a risk model developer look like then?
The work delivered during a typical week is a mix of incremental steps towards larger future goals. A long-term goal may be to have an updated model calibration ready for a capital review exercise planned in several months. For this purpose you may need, for example, to complete a set of statistical tests by the end of the week to verify that the mathematical assumptions of your model's framework remain fulfilled on the new data. Then there are tasks that arise at short notice: a management committee has questions about how your model impacted a recent report, a regulator requests some information, or the model needs an unplanned update because the range of input values changed significantly. Alongside this work, your team will be looking into ways of improving its workflows, such as developing tools that make recurring tasks more efficient or adopting new IT systems. Within a single week you're likely to program in a language like Python or R, write reports or model documentation in Word or LaTeX, and sometimes make PowerPoint slides. Part of your work on quantifying risks may be of a more qualitative nature too, which means you may be meeting with people in different areas of the bank and compute projections based on qualitative judgment with their help.
“I clearly felt how the analytical thinking imprinted by eight years of physics helped me at work, guiding my intuition on how to dig into a dataset or telling me which modeling approach may be worth pursuing.”Szymon Hennel, D-PHYS alumnus
Tasks that span longer time scales are the update of existing models (in view of new data or revised regulatory expectations, say) and the building of new models. For example, regulators have recently begun requiring banks to incorporate climate change in their capital management: this creates the need for a whole new class of financial models.
How did your physics studies prepare you for your job at UBS? From a professional standpoint, what do you value most of your background?
Excellent analytical skills are the most directly applicable. Physics makes you think in structures and concepts: groups, representations, components, symmetries, criticality, scale invariance, etc. This way of thinking is transferable to any other complex system with interactions. I clearly felt that the analytical thinking imprinted by eight years of physics helped me at work, guiding my intuition on how to dig into a dataset or telling me which analysis or modeling approach may provide useful results and be worth pursuing. Another important skill is general data analysis. Python and R are commonly used in the risk modeling industry. My experience of working with experimental data (choosing data containers, adopting different visualisation methods, cleaning data with a focus on critically assessing potential outliers, and so on) could be applied directly to modeling financial data. In this sense my work at UBS felt familiar from day one. As treasurer and then president of the Scientific Staff Association at D-PHYS, I also gained some experience in stakeholder management, which is an important aspect in the private sector.
A crucial aspect is that you learn to learn. To get up to speed on a variety of topics you don't just need intellectual ability: what helps is a learned methodology for information searching and prioritisation. In the corporate world, this approach helps you work as an independent employee who can take on responsibilities, making good judgment calls on when others need to be involved.