JPMorgan is an enormous employer of technology talent. The bank employs around 256,000 people globally, some 20 percent of which work in technology roles. Speaking at the recent AWS re:Invent conference, Rahul Arya, JPMorgan’s head of global technology architecture, said the bank employs more than 50,000 technologists in total, working on over 6,000 applications.
As JPMorgan seeks to get value for money from its technologists (and from its massive $11 billion technology budget), it’s placing more emphasis on creating reusable products. Daniel Pinto, chief executive of the investment bank, has long advocated for a technology transformation. “We need to avoid building everything from zero to one hundred every single time,” Pinto said at the bank’s investor day in 2019.
At re:Invent, Arya and James Reid, JPMorgan’s head of engineering and architecture, explained how the bank has set about achieving this goal. Specifically, the firm has used modular design to create “opinionated blueprint architectures as products,” said Reid.
It’s built a compliant and secure blueprint architecture in the cloud that adheres to JPMorgan’s compliance and security policies. The resulting architecture enables developers to self-serve from the blueprints available and to create new ones. Developers have autonomy and efficiency within the bank’s predefined parameters.
It’s a bit like when you buy an apartment versus a house, Reid explained. In a house you have the freedom to knock down walls and to do what you want, but in an apartment you expect to find key products such as stoves and dishwashers already installed and vetted for safety.
In this metaphor, the blueprint model is the apartment; it’s based around modular blueprints that can be reused and which have security built-in. Blueprints are effectively products in their own right, Reid said. They’re like Lego to the extent that they slot together different parts of JPMorgan’s layered model to create different blueprints for different purposes.
The blueprint model has five layers: A layer comprised of users, groups and encryption; a layer for external connectivity; a networking layer; a layer with a data lake (and related tools like machine learning), micro-services and web apps; and a final layer of containers and apps.
“If we discover a bug, we make a change once, republish it on the model registry and have all the teams upgrade to the new version,” Reid said, adding that JPMorgan now uses “GitOps” principles, version control, collaboration and compliance to move much more rapidly than in the past. The bank has been able to “scale up” its migration to the cloud as a result.
Reid had a word of caution for other banks attempting to achieve the same results. When JPMorgan first began moving to the cloud, he said, the bank allowed engineers to build and cull cloud-based products. However, this led to “death by a thousand cuts”: In an organization of massive size and complexity, it’s better to focus on a layered model based around blueprints in a center of excellence to avoid “friction between teams.” Accordingly, JPMorgan has a “cloud council” (headed by Arya) which provides guidance on the enhancement of blueprints and approves new ones in line with the bank’s strategy and guidelines.
A modified version of this article originally appeared in eFinancialCareers.