XiNG provides the platform for Citi’s growth

Shubham Dhage/Unsplash

A common standards framework and a cloud-native execution platform have unified pricing, risk and analytics across Citi Markets, without replacing the quant libraries the business already trusted. In an exclusive interview with Risk.net, senior leaders from Citi’s Markets Quantitative Analysis (MQA) and Technology teams explain how

Most technology strategies are built around the requirements a financial institution has today. The architects behind Citi’s enterprise technology risk management platforms, Xi Next Generation (XiNG) and XiP (XiNG Platform), set out to build for those requirements that have not emerged yet.

Pricing models, quant libraries and risk infrastructure are multiyear investments for any large bank, and they tend to evolve independently inside each business. The conventional remedy for the resulting fragmentation is a transformation programme to retire legacy applications, rebuild libraries and migrate everything onto a single platform. It promises simplification but carries significant delivery risk.

Citi’s leaders chose a different path. Rather than replacing mature quant libraries, they built XiNG – a common set of standards governing those libraries, and the foundation for Citi being named Risk’s 2026 Derivatives house of the year – alongside XiP, the cloud-native platform that executes against them. Together, they represent an ultra-modern and scalable risk management platform. What began inside the fixed income business now spans pricing, risk and analytics across Citi’s Markets division and the wider firm.

In this interview with Risk.net, Youssef Elouerkhaoui, head of MQA; David Tonge, head of quant development in MQA; Dominic Burgan, head of the XiP; and Nikhil Joshi, chief information officer for Markets, explore why standardisation, rather than replacement, became the foundation of Citi’s markets technology strategy.

Transforming quantitative architecture

Youssef Elouerkhaoui, Citi

The technology at the foundation of Citi’s risk and capital management system was laid more than a decade ago and began with exposing the analytics through Microsoft Excel add-ins, Elouerkhaoui recalls. “Whether you’re calling the library from Excel, from a Java-based system, from a .NET system or, more recently, from a Rust-based system, it’s all going through the same keyhole interface. You’re not translating back and forth; you’re not losing information.”

Xi originally served as the interface that exposed Citi’s models to the different programming languages used to write its applications. As adoption grew, Citi’s quants developed a strongly typed implementation to improve debugging and provide the consistency needed for production-scale deployment. The result was XiNG.

The technology platform standardises how users and systems interact with models, as well as how a valuation is requested, and how a stress test is applied to market data regardless of which library is relied upon.

The result is a federated architecture rather than a monolithic one, which significantly reduces the operational and delivery risks traditionally associated with enterprise transformation.

XiNG itself is a set of independent libraries, and they speak the same language, says Elouerkhaoui. “That allows the businesses to move at their own speed. You’re not joined at the hip, but you can move at your own speed,” he explains. “But then, when you need to do anything across the enterprise, you’re doing that very consistently.”

Critically, nothing was discarded, and existing libraries remained in production while gradually being adapted to operate through XiNG. That avoided the reconciliation challenge.

David Tonge, Citi

“We didn’t throw away our own quant libraries. We didn’t say that those are the heritage legacy quant libraries and we’ll start some fresh ones over here,” says Tonge. “These are the quant libraries, we’re fully invested in them, and they’re the best quant libraries on the Street.”

Migration happened one capability at a time, from valuation, internal stress-testing, independent price verification to risk management, while libraries provided a solid foundation to it all.

Building for scale

Scalability was a guiding principle from the outset, Citi’s panel highlighted.

While XiNG provides the standards layer, XiP delivers the execution environment. Designed as a cloud-native platform, XiP provides the infrastructure required to execute pricing, risk and analytics consistently across the front office, risk and finance.

One of the platform’s distinguishing characteristics is its application programming interface (API) architecture, which standardises how enterprise workloads are expressed regardless of whether they originate in trading, risk management or finance.

Dominic Burgan, Citi

“XiP creates a unified foundation for calculations across Citi Markets, moving us from a variety of specialised systems to a single, consistent language,” says Burgan.

Users now describe what they want through one API, specifying trades, environment, analytics and measures. Full-revaluation stress tests run on demand during peak volatility, because cloud-native compute is elastic.

“We’re not just doing existing things faster,” Burgan says. “We’re doing things that simply weren’t possible before.”

Citi’s peers have built similar platforms, such as Goldman Sachs’ SecDB and JP Morgan’s Athena.

What differentiates Citi’s approach from the rest is that adoption is not ‘all or nothing’. Each asset class takes what it needs, from the XiNG standards as a minimum, up to the orchestrator, calculation services, risk stores and public cloud.

“It can be customary for the industry to take an all-or-nothing, or a one-size-fits-all, approach when it comes to large risk platforms,” says Joshi. “And, with XiP, we’ve taken quite the opposite view. This gives what I like to call ‘freedom within the framework’: the framework is defined, but asset classes have the freedom to navigate depending on where they are.”

A key innovation of XiP is the standardisation of Citi’s core financial data and calculations. Now, the same trusted figure is used across the front office, finance and risk departments, significantly boosting efficiency and regulatory consistency.

Business integration to enterprise architecture

The platform’s implementation is best demonstrated by the value it brings across the capital management function.

“XiP has transformed our approach to capital management by enabling daily risk-weighted asset [RWA] calculations for every trade, desk and business unit,” Burgan says. “That shift has turned capital from something that was abstract and quarterly into something the business can actually manage in real time.”

That cadence underpins Citi’s RWA auction mechanism, in which businesses trade unused capital capacity with one another within the quarter.

While XiP did not do that alone, Burgan emphasises that “there were lots of technology and quant teams collaborating to solve this. But what XiP provided was the foundation that made the disciplined dynamic capital allocation possible at scale.”

The second case study comes from the second line, which typically operates beyond silos. One system has run internal stress-testing for the whole of Citi Markets since 2021 and, in 2025, it was used for Comprehensive Capital Analysis and Review, reducing cycle time by a factor of 10.

Nikhil Joshi, Citi

One platform and one compute layer now also serve present value, risk, profit-and-loss (P&L) attribution and stress-testing, end of day and intraday.

Citi has upgraded its trading and risk management capabilities with XiP, moving from a periodic view to a live, precise assessment of risk and P&L. While a periodic view might have been sufficient in less volatile markets, it limited the ability to accurately assess a derivatives book and monetise its gamma during periods of heightened market movement.

During recent periods of market volatility – led by geopolitical tensions and erratic moves in oil and other commodities – Citi’s teams were able to generate firm-wide stress tests in near-real time, by individual factor or in aggregate.

“Previously, when we’ve had periods of volatility like this … the question was around how long this would take. Would it take a few weeks? A few months?” says Joshi. “Now the questions are: How many minutes? How are we going to run it? Who needs to see it? It’s become a little bit of a lingua franca.”

From spread products to enterprise adoption

XiNG’s first front-office implementation followed Citi’s integration of its credit, mortgages and municipal bond businesses.

Rather than rewriting multiple quant libraries into a single platform, XiNG’s architects proposed retaining the existing libraries and market data. As a result, spread products became the first business to run on XiP, providing a single platform for risk and P&L attribution.

“It is very important at the people level and the organisational level to be joined at the hip. So either you win together or you lose together,” says Joshi.

Governance responsibilities remain unchanged, but the design and development of XiNG and XiP have been closely co-ordinated across MQA and Technology teams.

Standards are developed through joint working groups with technology representation on every committee, combining technical expertise with broad stakeholder engagement. The Citi panel emphasised that the collaborative approach has helped drive adoption, supported by sustained executive sponsorship over several years, rather than a series of short-term initiatives.

New frontiers for infrastructure and innovation

While XiNG was initially conceived to address consistency and governance across model libraries, its architects see its longer-term value in enabling the next generation of quantitative innovation.

One immediate application is Citi’s work on the Fundamental Review of the Trading Book internal models approach, where large-scale computation, complex modelling and enterprise-wide consistency must come together.

“The opportunity to have one consistent platform for the whole thing just removes complexity,” says Tonge. “Citi has grown as an integrated bank, and XiNG is a big part of that. It makes it easy for us to think about the business as a whole.”

As the core platform matures, attention is increasingly turning towards engineering optimisation rather than infrastructure buildout. Future priorities include improving cloud efficiency, giving users greater self-service capabilities within controlled guard-rails, and supporting more demanding workloads such as near-real-time risk analytics without significantly increasing compute costs.

The platform has also been designed to lower barriers for researchers and quantitative teams. Rather than prescribing specific use cases, XiP provides a shared foundation that allows users to experiment with new analytics while maintaining appropriate governance and controls.

“XiNG is a story about innovation, business impact … and a very strong partnership with our technology and business partners,” Elouerkhaoui says.

That adaptability may ultimately prove to be XiNG’s greatest strength. Rather than solving a single modelling challenge, the platform establishes common standards that allow new analytical capabilities, regulatory requirements and business priorities to be incorporated without repeatedly redesigning the underlying architecture.

In an environment where market, regulatory and technology demands continue to evolve, that ability to adapt at scale may become as valuable as any individual application.

Sponsored content