We’re inviting applications and are available to answer your questions on the program and on AI in financial services and in markets roles in particular. The program will cover how to use AI in markets and risk jobs, and how to structure alternative data for investment decisions.
Let me start by question (1) : Europe is a great place to work on AI projects. Think about the come back of Yann LeCun in France ; or all the meta, deep mind, etc having a research center in Paris ; or have a look at the Alan Turing Institute on the UK and the great research that is done at Oxford on the ethics of algorithms.
But since we are talking about financial markets and quantitative investment: big names are there too, Europe in unavoidable for this business: Amundi, Allianz, BNP Paribas, Ardian, for Europe-based ones, but all global financial companies have a presence in Europe, and in most cases a R&D team because the talents are there.
Working in AI for Markets and Quant Investment means to work in a regulated environment, i.e. building more industrial and long term solutions, and Europe is good for that.
[EDIT] Let me add a link to this linkedin post “CFM + AMI“.
hello @Hoid2534 the main difference between financial markets/quant investment vs. the other domains is that there is no repeating experiment: in general, if you monitor any recording of a sensor (temperature, pressure, electricity, biological, or even logs of computers, etc) this is more a less an “almost repeating” sequence of events or segments of regime that are very well defined. I am not saying it is easy, but if you cut a long time series in time windows, an algo that is nonlinear enough (meaning: it will contextualise) and has enough data (in-sample v out-of-sample, etc) will succeed in capturing something.
In the financial and economic world, we are observing only one “omega” (one realisation) in an economic context and business cycles that evolve. As a consequence the signal over noise ratio is really low.
To get familiar with such a noisy context, you should to some kaggle competitions, or look in this data challenge, and look for datasets coming from quantitative investment and financial markets.
In terms of knowledge, you should look in one of these three directions:
for the Master AI for Markets and Quantitative Investment you need a good level of maths, because we will teach how to change the structure of the algorithm to cope with the specificities of financial market/investment data. My answer to @Hoid2534 gives you some examples. But let me list more
AI/ML uses Reinforcement Learning (RL) and Markets and Investments use Hamilton-Jacobi-Bellman (HJB). One the one hand you have Q-function, and on the other hand Value-functions: how are they related? you need some maths (in particular the Dynamic Programming Principle, and hence you need to understand Markovianity) to be at the crossing of RL and HJB.
in AI/ML you can to a lot of high-dim compression (looking for low rank + sparsity), but if you want to do that on time series of financial returns + liquidity, you will face some long memory effects that will need to rethink compression. => you need maths
Covariate shift is a well-known effect in AI/ML, but how is it related to “regime switching” in macroeconomics (with the BVAR model for instance, to be compared to LSTM or Kalman filters)? again you need to go back to the maths.
Hi to the both of you. As a recent bachelors graduate weighing up whether to go back to school or straight into work, my question is what are the benefits of doing a masters today? The job market is extremely tough at the moment and it feels like companies have a preference for people with experience rather than education at the moment.
Your course is directly related to my career path and looks very interesting, but I’m worried about falling behind my peers going into full-time work. What kinds of internship prospects are there for masters students?
Thank you for doing this. I just wanted to ask - which careers in markets and FS more broadly do you see as most “AI-proof”? Which one will either never be or could never be automated and forgotten about?
In general I would recommend to do a Master (and then a PhD!) when the job market is saturated, it is a good way to capitalise knowledge for your future career. Unfortunately, what you feel like having a preference for “people with experience rather than education“ is just a way to say “freeze recruitment”. The same companies will open internships, and you will learn more during these internships.
The discussions we have with the partners of the Master program suggest they will provide interesting internships to our students. Even outside of the partners, I see internship position open for students skilled in a good mix of ML and Quant Finance. I have no concern for the students of this program.
hello @Estimation this is a funny question: the master we are putting in place is exactly “AI-proof” in the sense the students will learn how to solve quant finance/market problems with IA, and even better: to tune and adjust or innivate AI/ML solutions that for.
But if you really want to discuss this idea of “future-proof jobs” in financial markets and systematic investment, I would say that tasks are automated, not jobs. Jobs are serving a purpose: maintaining a feature at the core of the company you are working in.
Then you have to go back to the root: what are the functions of the financial system?
have a look at the great paper Merton, Robert C. “A functional perspective of financial intermediation.” Financial management (1995): 23-41. and you will see that Merton is listing six features:
A financial system provides a system for the exchange of goods and services.
A financial system provides a mechanism for the pooling of funds to undertake large-scale indivisible enterprises.
A financial system provides a way to transfer economic resources through time and across geographic regions and industries.
A financial system provides a way to manage uncertainty and control risk.
A financial system provides price information that helps coordinate decentralized decision-making in various sectors of the economy.
A financial system provides a way to deal with asymmetric information and incentive problems when one party to a financial transaction has information that the other party does not.
Of course any company is not implementing all these features, sometimes only one (for instance agency brokers focus on 6) or a handful (asset managers are providing 3+4+5).
If you are aware that your job is not a list of tasks, but if you focus on understanding how you help your company to answer to a core question, then you will never be replaced by an IA.
In the MaQI program we will teach techniques, tools (ie how to solve tasks), but also why they work (to enable student to modify them, ie modify the tasks), and more importantly we will have sessions on the business models, to help student to get aware of what is the purpose of their job.
I’m experimenting with some machine learning projects at the moment and I’m not sure whether to use TensorFlow or PyTorch. Which do you think stands out to employers more, and which is used most widely in the markets?
Thank you for posting here. I want to ask - I graduated three/four years ago and didn’t really do great in my exams. But I managed to make some connections at school and I’ve been working in the industry (on the dev side) since then - I’m not sure if the program is necessarily targeted at me but I think it would be 1000000x what an MBA could do for me or my career.
Are you guys accepting mature students for this - if not, would you considering hosting an MBA-type degree, executive or otherwise, specifically for mature students?
Hello and thank you for this. Can you say what kind of background you’re looking for for this course? Is it not better to study a broad masters in finance? It seems like it might be a mistake to specialise so early
Thank you for sharing MaQI — it’s encouraging to see a program that treats AI in markets as a real operational and risk tool rather than a purely academic topic.
I am currently working on a real-world financial forensics case involving crypto exchanges, custodial wallets and cross-border asset tracing, combining blockchain data, platform records, regulatory filings and court material into a unique applied dataset of how modern market infrastructure behaves when things go wrong.
If helpful, I would be glad to contribute this as an applied AI use-case for MaQI, particularly around alternative data engineering, anomaly detection, custody risk and AI-assisted reconstruction of financial events.
it is a matter of choice: if your goal is to work at the interface of AI and financial markets, it is better to link the concepts as soon as possible in your brain.
An example: derivative pricing; if you follow a standard MFE, you will go through the definition of derivatives (same in the MaQI M1), then stochastic control (same in the MaQI M1), and after that you will attack the real problem: value function, martingales and conditional expectation, dynamic programming principle (under Markov assumptions), and Hamilton-Jacobi-Bellman (that is a PDE with a sup/inf). After that you will learn the classic PDE solver methods (Euler schemes, etc).
This is where MaQI is different, at each of this step you will have the parallel with Reinforcement Learning, from the starting point (Q-function in place of Value function, to teach the difference), to the end (Offline RL in place of Euler, to see the difference, and they are some!). Our M2 courses are jointly taught by 2 professors, one from ML/AI, one from Quant Finance.
you need to be good in applied maths / stat / proba, and be ready to learn basics in economics. We do not have any specific issue in accepting mature students, even if the vast majority will be of “standard age”, diversity is good, of all sorts.
but we have issue in accepting students who will not make it in terms of technicality, because we will teach you the theoretical reasons why it works (or not), and give you the opportunity to try (in python on real data).
Hello Charles. I note that you have worked in quant roles in banks and on the buy-side. Can I ask you a non-Masters related question about your experiences in these jobs. Which did you prefer and why? And do you think that data analysis jobs are going to disappear as AI becomes more used?
Thank you, Professor Lehalle — I appreciate the openness.
I will reach out to you directly with a short overview of the case and the type of applied AI questions it raises, so we can see whether it fits well with MaQI’s teaching and research objectives.
