Stage 3A - Ingénieur quantitatif - Reinforcement Learning & Local Volatility Calibration H/F

Job ID:  39482
Location:  PARIS (FRA)


Murex is a global fintech leader in trading, risk management, and processing solutions for capital markets. Operating from our 19 offices, 2,500 Murexians from over 60 different nationalities ensure the development, implementation, and support of the MX.3 platform which is used by banks, asset managers, corporations and utilities, across the world.

Join Murex and work on the challenges of an industry at the forefront of innovation and thrive in a people-centric environment.
You’ll be part of one global team where you can learn fast and stay true to yourself.



Team :


The MACS (Murex AnalytiCS) modelling team is responsible for the implementation of efficient and innovative evaluation methods and the computation of risk measures (Sensitivity computations, VaR, PFE, XVA...) for financial products (from vanillas to the most exotic ones). It is a cross assets team (Equities, FX, Rates, Commodities and Credit) which understands, implements (calibration and financial evaluation) and maintains standard modelizations as well as more innovative ones to fit market needs.

MACS models can be used through a pricing library but are also exposed using a REST service. Particular care is put on producing accurate and reliable results in a timely manner, which leverages on modern technologies (ex: GPU computing) or adapted numerical methods.


Missions :


Machine learning / Artificial Intelligence methods are increasingly part of our lives today, just like in financial markets.
First “Supervised Learning” technics using deep neural networks were put in place to mainly approximate very accurately and very quickly valorization functions (complex and time consuming) of specific financial products (vanillas as well as exotic ones) under several modelizations. “Unsupervised Learning” technics were also used to develop, for example, new valorization methods, alternatives to PDE resolutions or Monte Carlo methods. The next step is then the use of “Reinforcement Learning” technics.
The purpose of this internship is then to implement “Reinforcement Learning” technics in the context of the calibration of financial models. In fact, these methods are designed to learn throughout the diffusion process and the different simulations the parameter values of the model so to match market option prices.
In a first step of this internship you will familiarize yourself with the “Reinforcement Learning” technics. A first implementation will be developed under a simple modelization (Black & Scholes) to facilitate on a low dimension context a first analysis of the method and determination of the main drivers (simulation number, discretization steps, reward policies…).
Next step is to implement these technics to more complex modelizations (the local volatility model will be the first one) but still focusing on matching the market prices of European options. You will then adapt the method so to match also more exotic financial products prices (i.e. one touch options / VIX options...).
A particular focus will be put on the quality of the implementation and analysis as well as on the computation time.


Profile :


3rd year student in Computer Science / Financial engineering or Master Student in Mathematical Finance with:


  • Strong knowledge of quantitative finance
  • Good knowledge of C++ (knowledge of Python or of parallel computing would be a plus)
  • Interest in financial markets
  • Interest in software engineering, continuous integration, devops, etc...
  • French and english speaking
  • Ability to work in an agile and highly collaborative context