In conversation with Xabier Anduaga: How Open-Source Is changing the quant landscape

As the traded risk environment continues to evolve, Xabier Anduaga, Partner, Quantitative Services at Post Trade Solutions, LSEG, discusses the current challenges in this space and the growing interest in open-source frameworks like the Open-Source Risk Engine (ORE).

In a period of profound disruption, the challenge of distinguishing signals from noise has never been more critical. To address this, there is a growing need for integral risk solutions systems, data, and models that can process information quickly and help decision-makers act in a timely manner. 

These challenges grow more profound as client priorities evolve. For example, intraday risk is becoming increasingly important for many firms. The nature of stress testing has also evolved; it is no longer just a regulatory checkbox, but a core part of the risk framework. 

Meanwhile, for banks looking to adopt the new Internal Models Approach (IMA) under Fundamental Review of the Trading Book (FRTB), a full overhaul of their market risk infrastructure may be required to accommodate new analytics and requirements at the trading desk level (e.g., PLAT, RFET). Many firms still rely on legacy systems that weren’t built with these requirements in mind, which creates significant pressure on cost.

Revamping market risk or Counterparty Credit Risk (CCR) infrastructure is no small task, and there are often many challenges to overcome. 

These might include dealing with outdated system design or architecture that cannot easily accommodate new requirements, as well as fragmentation across asset class or risk type, which can lead to inconsistent data models and analytics, making the integration required to deal with increased demands difficult. Firms may also face knowledge transfer and key-man risks where only a small group of quant developers build and maintain critical infrastructure, leaving a knowledge gap internally when they depart from the firm. 

For many firms, the only viable path forward is to replace their infrastructure entirely.

This is where open-source solutions like the Open-Source Risk Engine (ORE) are seeing major growth. During FY24 alone, we have seen the number of institutions actively using ORE double, and the pipeline of institutions interested in adopting ORE increase fivefold.

ORE offers a cost-effective alternative (i.e., no license fees) to in-house builds or commercial vendors. For some firms adopting ORE at scale, this has translated in multimillion pounds per year in savings. On top of this, the firms actively using ORE get benefits that go well beyond cost, including:

• Consistency – ORE offers a unified platform across asset classes and risk types where trade data, configuration data, market data, and analytics are uniformly represented.
• Transparency – There are no black boxes, which simplifies auditability and model risk assessments, giving users access to extensive detailed documentation with 100+ examples.
• Active research and development (R&D) – Unlike analytics vendors that often lose their incentive to continue investing in R&D, e.g., when acquired by private equity firms, ORE users benefit from community-led development, which incorporates funded development requests as well as development done as part of feedback obtained through our ongoing support to active ORE users. For instance, our most recent ORE releases include new, cutting-edge features like Adjoint Algorithmic Differentiation (AAD) and GPU acceleration for high-performance exposure simulations.
• Time to market – Implementation times and version updates are usually faster than third-party solutions.
• Proven performance –our clients globally rely on ORE and Post Trade Solutions for business-critical processes such as daily ISDA SIMM™ calculations as well as Initial Margin backtesting and benchmarking. This is key to note, as it helps challenge the outdated perception that open source lacks the resilience or credibility required for risk management at scale.

ORE is also well-suited to support model risk functions in regulated environments, as it offers a robust set of analytics for benchmarking internal models and validating third-party tools.

The main thing to consider is the implementation model. Many firms prefer a locally hosted setup to retain full control. For this approach, getting buy-in from the quants is essential. It might sound trivial, but their support is critical to a successful ORE journey. Local implementation is not just about control, though; it also gives firms the flexibility to build ORE extensions for features or models that they see as a competitive edge and prefer not to share with the wider community.

It’s worth noting that while a local ORE implementation is much simpler than a full in-house build, some firms will still need to consider if they have the adequate resources to deploy, host, and maintain once implemented at scale. 

As an alternative, Post Trade Solutions has launched a SaaS version of ORE called Risk Analytics Lab. It offers the same transparency, cost-effectiveness, and community-led development benefits as ORE, but without the complexities of local deployment. Risk Analytics Lab is hosted alongside Post Trade Solutions’ well-established risk services, uses the same centralised market data and hosted configurations, and allows users to customise how they interact with and operate ORE, for example, through hosted Jupyter notebooks.

As with any new technology, the journey usually starts with a proof of concept, adopting ORE for a specific use case on a limited set of trades or portfolios, for example, as an XVA engine. At this stage, ORE is configured to access market and trade data, enabling testing and benchmarking.

Once value is demonstrated, the next step is scaling up, expanding to the full trade population and risk factors for that use case. This phase focuses on performance and scalability.

By this point, the benefits are already tangible for the client. The natural next step is to extend ORE to other use cases. For instance, if ORE has been implemented for XVAs, why not leverage the same exposure engine for PFE calculations and benefit from the large computational savings of a unified exposure engine? Once the data is in place, switching the PFE module is largely a matter of configuration. 

From there, it’s just about maintenance, monitoring performance, and keeping ORE up to date as new versions are released.

Clients see real value in having us involved during the implementation phase. Fully understanding the complexities of ORE can take time, and working with our quant experts accelerates the process and helps avoid common pitfalls. Our support often includes developing interfaces for trade data and market data, assisting ORE configuration, and reconciling exposures and sensitivities. Our quant consultants work alongside our ORE development team to maximise efficiency, prioritizing critical items and reducing delivery timelines when compared against alternative delivery models i.e. independent consultants or third-party service providers.

Over time, it’s not uncommon for our clients to become ORE experts themselves; some even contribute to the open-source code, which we’re always proud to see and actively encourage.

Post-implementation, we continue to support clients on an ad-hoc basis, whether it’s troubleshooting, interpreting model outputs, or expanding ORE’s use. We have an SLA framework in place to facilitate this.

Beyond ORE, we support clients in developing, enhancing, or validating their own model suites. Our quants have extensive experience working with both the sell-side and the buy-side, which positions us well to work with a broad range of institutions.