Today, Paul Bilokon (Thalesians Ltd) interviewed Dariush Mirfendereski, an alumni of the Machine Learning Institute, established by World Business Strategies, Thalesians, and Thalesians Marine. For many years, Dariush used to be Global Head of Inflation Trading at HSBC.
Paul Bilokon: At the time when you were studying at MLI, you were Managing Director, Global Head of Inflation Trading at HSBC. You have worked in this role for over a decade. What can you tell me about this role? What were the upsides and challenges?
Dariush Mirfendereski: I headed a trading desk, with a similar global role from 2004 through until 2022 (UBS then HSBC). I led a trading team and also had trading responsibilities myself. The tasks of the desk were principally market-making and managing risk, serving customers on the institutional investor side (real money and fast money) as well as sovereign wealth funds and large corporates using inflation-indexed instruments, both sovereign bonds and inflation-indexed derivatives. This was across the main developed markets (UK, Eurozone, US), with some additional focus on markets in Japan and Australia (at UBS), New Zealand (at HSBC).
PB: You completed your PhD at University of California, Berkeley. Your thesis was entitled “Probabilistic Characterization and Response Prediction of Micro-electro-mechanical Systems (MEMS).” How useful was your PhD work in your career, and how have things changed now, with the emergence of AI?
DM: The PhD led to an engineering job in California. When I later transitioned to a job in banking, the PhD was useful as a technical credential when being hired for a role on a structured rates trading desk where many traders are typically former quants and/or come from a very quantitative background. The PhD was also useful for giving me the confidence to solve novel problems using quantitative techniques and doing my own research. AI/ML has now been incorporated into many of those quantitative aspects (this did not exist in 1996 when I started in banking).
PB: Since you have graduated from MLI, you decided to pursue machine learning full time and have joined UCL‘s Master of Science in Machine Learning. How did your MLI certification prepare you for your full-time studies and research in ML?
DM: I felt that after the 6-month part time MLI course I needed more knowledge, hence pursuing the ML MSc. MLI certification was an excellent window into what was possible for applying ML to finance. In some ways I wish I had done MLI after my MSc when I would have been more proficient at Python and more conversant with ML methods in general so that seeing real-world applications in finance would have then been the ‘cherry on top’, allowing me to learn/ask a lot more than I did when I took the MLI course. I may well choose to do the MLI course again in the future.
PB: What do you particularly like about the MLI approach to teaching ML/AI?
DM: The live coding and practical approach and examples were extremely appealing. Of course, what you get out is linked to what you put in. Given 50-60 hour work weeks, sometimes doing the assignments and preparations prior to the evening lectures was challenging. I would highly recommend leaving aside time during the weekend to prep for the week ahead in order to get even more out of the MLI course.
PB: At UCL, DeepMind has a significant presence, several scientists from DeepMind teach courses at UCL. Did you attend any of them? What can you say about your interactions with DeepMind?
DM: DeepMind staff taught the RL course. To my knowledge that was the limit of their teaching interaction in the ML MSc. There are weekly, informal, research oriented seminars and DeepMind staff would present at some of those. We also had the occasional DeepMind presentation at the Computing Society events. Finally, students may choose to do their MSc project with DeepMind linked professors.
PB: You are now working as a freelance consultant to major financial institutions. How is your current work different from your previous full-time work? What are the pluses and minuses?
DM: I have actually accepted a new role back in banking which I will start later this month. It will be a Managing Director role trading inflation. I look forward to also applying ML techniques in my new role.
PB: What can you say about the impact of ML/AI on inflation markets?
DM: A large impact has been on inflation forecasting methods. These, of course, have applications across all markets, not just the inflation markets. There are several firms now using AI/ML in inflation forecasting (e.g. Turnleaf Analytics, Alternative Macro Signals). These are in the ‘public domain’, however, many banks and hedge funds also use their own in-house ML-based methods for inflation forecasting. There are also a range of applications on the quantitative side which can help in achieving better risk management, finding RV opportunities, etc.— areas where traditional quantitative methods are enhanced with ML methods. These apply to the full range of rates/credit/equity/FX markets but, with the help of strong domain knowledge, can also be appropriately applied to inflation markets.