Team & Context

XVA calculations module is responsible for scientific computations of financial risk measures, mostly XVAs and XVA Greeks.

These computations are delivered:

• as large reports (nightly runs, sometimes requiring hours of computations),
  requiring optimizing in the large (esp. IOPS management),

• as well as incremental computations (happening during daily activity of banks),
  requiring good responsiveness in concurrent usages.

Elaborated Risk computations, supported by a Domain Specific Language

Computing xVAs from Monte-Carlo generated Grids of stochastic Market Values is an interesting matter.
The biggest difficulty comes from the variety of configurable outputs that the XVA module supports. This matter is addressed thanks to Domain Specific Language, created and owned by the team to handle easily complex configurations, while enabling non-intrusive execution parameterization (sequencing, distributing).

Things get more subtle with What-If and Sensitivity computations: How to re-compute just what is needed for incremental computations, propagating correctly and minimally the impacts?

Distributed Computations (On-Premises and Cloud) with Challenging Non-Functional Requirements

Operating similarly to MapReduce algorithm on Graph-structured computations, the XVA module…

… orchestrates per-trade computations on computation nodes (which generates large data volumes), … that are then “aggregated” (reduced, in MapReduce’s terminology) into XVAs, using concurrent computations (multi-threaded and/or multi-processes).

The NFRs of the Aggregation computations have been evolving in the past years with:

• the introduction of XVA Greeks (which are portfolio level Greeks) computations during trading hours,

• the needs of clients for fine grained time discretization,

• the evolution of the available hardware, being On-Premises or Cloud (e.g. new generation of SSDs).

It is a constant challenge to deliver XVA computations with performances compatible with our clients’ XVA desk activities.

Your missions

• Development XVA financial computations, in a functionally rich context:
  the Murex platform handles a large variety of Market Data and financial products.
  Developers and testers collaborate using a BDD workflow.

• Integration of XVA computations in the current system, which handles large volumes and concurrent usages, and consumes large inputs (retrieved by Java RMI or REST APIs).
  Specific attention is dedicated to Observability and Scalability, since the system is distributed as well as cloud deployed.

• System analysis to solve reported bugs and improve the NFRs.
  Most of bug investigations and performance analysis are performed by analyzing the data produced by the system: structured and correlated logs are analyzed with utilities owned by the team, all intermediate computations are dumped, etc.

Your profile

Computer Science Degree, with skills in:

• Java development

• Unit testing

• Refactoring

Interest for:

• developing financial computations,

• testing performances and optimizing in the large,

• domain specific languages.

Everyday work in English, since half of the team is international.