Isabella Velásquez
Industry

Embracing open-source data science for smarter financial risk decisions

Data science is a cornerstone of risk management in finance, with banks continuously investing in advanced analytical tools and infrastructure to navigate complex markets and regulations. As the financial landscape introduces novel challenges, from new regulations to the transformative applications of Artificial Intelligence (AI), the advantages of open-source tools become more evident than ever.

The advantages are both technological adoption of algorithms and infrastructure and the empowerment of the talented professionals who build, interpret, and act upon these insights. This blog post explores the strategic embrace of open-source data science in financial risk management by drawing directly from the experiences and expertise of practitioners who have shared their valuable insights on the Data Science Hangout, our weekly event where we invite industry leaders to share what it means to be responsible, effective, and inspiring.

We’ve compiled insights directly from these leaders on how they empower risk teams to deconstruct opaque systems, implement rapid iteration, and harness open-source tools for their organization’s greatest strategic advantage.

 

Deconstructing the “black box” and building your own

 

Proprietary risk models offer powerful capabilities, and the allure of “throwing data in and getting an output” is strong. But professionals in financial risk demand a deeper understanding. As Kshitij Srivastava at Milliman describes risk analysis when launching a new type of product, “our usage is more on the traditional statistical side rather than sort of more machine learning oriented because, in general, actuaries typically want to understand sort of the drivers of risk and put a formula to it rather than sort of walking through a black box.” Analysts then gain the deep understanding needed to stand confidently behind their risk assessments.

 

 

The need for actionable insight extends beyond the quants to crucial business stakeholders. Having the ability to change an output and reassess risk immediately is incredibly powerful. Daren Eiri from Arrowhead General Insurance states, “It’s really just about being able to communicate to the stakeholders how those models work, what can they expect, what are the limitations of the model, and being able to do that effectively.”

 

 

Regulators, like those enforcing Basel accords or SR 11-7 in the US, increasingly demand model explainability and auditability. This is precisely where open-source languages like Python and R, and the extensive libraries built upon them, offer a critical advantage: full transparency. The entire logic is visible and available for rigorous internal validation and external scrutiny. Even well-understood models can degrade over time due to shifts in data distributions or relationships, a phenomenon known as “model drift”. Analysts can choose open-source tools, such as the vetiver framework, that incorporate built-in model deployment monitoring capabilities to detect and alert against these silent errors so models remain relevant and accurate.

This transparent, auditable nature of open-source tools allows financial institutions to not just “deconstruct the black box” but to build their own box: one that is fully understood, adaptable, and compliant with the stringent demands of modern finance.

 

Agility, customization, and innovation

 

Financial institutions can develop, test, and iterate new risk models with the help of open-source tools. This agility is vital for responding to emerging risks or quickly assessing new financial products.

Open source allows for the flexible creation of highly customized risk models that can precisely fit a bank’s unique data, portfolio characteristics, and specific business needs. This bespoke capability offers a significant competitive advantage, as analysts do not have to wait for someone to create or adjust a tool for them. Jason Foster at Marathon Asset Management leveraged the open-source nature of R to create his own package, roll, when the currently available options were not meeting his requirements.

 

 

The open-source ecosystem means that a bank’s risk analytics team is constantly benefiting from a worldwide pool of expertise. This agility is underscored by the rapid integration of cutting-edge advancements like AI. The open-source community quickly develops and shares new tools. For instance, ellmer for R and chatlas for Python are just two examples of tools that allow financial institutions to immediately incorporate powerful large language models (LLMs) into their daily risk workstreams.

 

Collaboration, talent, and the broader impact of the risk analyst

 

Open-source tools (like Quarto or version control with Git and GitHub) foster better collaboration among risk teams, ensure reproducibility of analyses (critical for auditing), and facilitate knowledge sharing. Michael Derstine at Wells Fargo describes how putting reporting and data analytics logic in flexible languages like Python or R and managing it in GitHub allows for incredibly quick cycles through development, user acceptance testing (UAT), and production:

 

 

Finally, open-source tools enable effective communication of risk insights to non-technical stakeholders through the creation of interactive dashboards and automated reports. This is crucial for making data science actionable. In Jason Foster’s Data Science Hangout session, he also highlights the common need for “custom tweaks” when calculating financial metrics, and he champions the use of tools like Shiny apps to allow users to toggle parameters and gain a dynamic understanding of their data.

 

The future of financial risk management is open

 

As financial institutions navigate an increasingly complex and rapidly evolving risk landscape, open-source data science delivers transparency, agility, and the power of collective intelligence. We’re excited to continue showcasing its profound reach and impact, and proud to be part of its ongoing evolution and innovation.

Tags: Finance