SS&C builds data mesh to unite acquired platforms

SS&C Technologies is renovating its in-house data management plan, and the company has decided a data mesh architecture is the way forward.

A data mesh is a socioeconomic structure that considers data a product, and considers business domains the owners of that product. The concept is a relatively new innovation in data management structures, as companies look to move away from traditional data initiatives, such as data lakes and data warehouses, due to the growing amount of data that financial services firms deal with.

SS&C’s data mesh platform, SS&C Everywhere, is a cloud-based data management and marketplace platform that was created in-house to simplify data access across multiple domains within SS&C.

Curt Burmeister, CTO of SS&C Algorithmics, explains that a mesh structure was needed due to SS&C’s growth through M&A. He says that as the company grew through acquisitions, it ended up with solutions using data that was siloed in different parts of the organization, which created problems. This sometimes led to multiple instances of the same software that were not consolidated, he says.

“This is a purely internal use case where, at SS&C, we have multiple Salesforce instances, where different business units have come into the company over the years and have different Salesforce instances,” Burmeister explains. “We have our own internal CRM system as well, so one of the projects we’re working on is being able to bring all that data together to help with our own internal reporting on customers and what are we doing with Client A versus Client B?”

There’s another [standardization] challenge, especially when you‘re dealing with different systems of matching up records, so this is one of the applications where we’ve applied some GenAI to the problem

Curt Burmeister, SS&C Algorithmics

Burmeister says that when the team at SS&C started this project a few years ago, while they were not the only team implementing a data mesh in the world, it “wasn’t very common,” but they’ve made significant progress since then. He says the team has connected the data mesh to the key systems that most SS&C clients use, but not yet all of the data sources due to the company’s size and client demand.

Following the footsteps

SS&C is not making this journey into the data mesh alone. Other companies have embarked on data mesh journeys and learned much from the process.

Previously the domain of smaller companies, data meshes are appealing to larger players, too. UBS Group CTO Rick Carey, who spoke to WatersTechnology in 2022, said that most traditional data management strategies resemble monolithic data architectures comprised of a number of disparate sources of data, which makes it hard for companies to provide data to customers.

“Data and consumers are increasing, and methodologies like AI, machine learning, analytics, computational capabilities, GPUs, and the cloud are also increasing. That to us is the sweet spot for a data mesh architecture,” Carey said. “Why do we think this? A data mesh architecture starts with the premise that data can be anywhere—and it already is everywhere.”

Lyndon Hedderley, head of customer solutions at Confluent, says the company decided to adopt the mesh architecture for its internal data management challenges, but large firms tend to have a tougher time.

“Since the 1950s, we’ve taken data from the sources, we’ve put it somewhere in a database, data lake or data warehouse, then we’ve built the application architecture on top of that,” Hedderley says. “I suppose a data mesh specifically is a socio-tech concept, not just a bit of software, so there’s a definite transformation required to implement and work with a data mesh, and because of that, the bigger and more complex the organization, the bigger and more complex a data mesh implementation will be.”

Hedderley says most large financial services firms are so used to managing their data in accordance with the ways of the past that implementing a new system across the entire organization and expecting it to work immediately is foolhardy.

“If you take a large bank as a huge enterprise with multiple business units, different ways of working, multiple project teams within business units, and come in and say, ‘Hey, let’s implement data mesh enterprise-wide,’ then that’s never going to fly,” he says. “You have to eat an elephant in bite-sized chunks.”

One of the chunks that Burmeister is keen for SS&C to swallow in its data mesh journey is the ability to standardize data across a range of data sources within the organization. He says that sometimes data from one company is split across multiple different names and account details, which gives the impression that there are multiple companies instead of just one. This is where generative AI comes in.

“There’s another [standardization] challenge, especially when you‘re dealing with different systems of matching up records, so this is one of the applications where we’ve applied some GenAI to the problem,” Burmeister says. “You could have a client in the system as JPMorgan spelled out like that, and in another system you might have JPMC, and in another system you might have J.P. Morgan, right? So you can have the same client represented or the same name represented in different systems. So what we‘ve done is developed some work around GenAI using vector embeddings.”

Vector embeddings are a way of converting words, sentences, and other data into specific numbers, which represent different data types. By using the example above, Curt and his team would be able to convert JP Morgan data into a specific value, and then put it through a large language model and run it against other data sources. If the other data values substantially match JP Morgan’s value, then it would be possible to match those records together and standardize the in-house data records.

Burmeister says that the vector embeddings method, which he calls “fuzzy matching” can also be applied to the company’s AI outfit, Blue Prism.

“If we look at Blue Prism, which is our Intelligent Automation Solution, we have solutions in the space that we call intelligent document processing,” he says. “So if you want to scan a bunch of leases and get information out on, say, the term of the lease, the ending date, etc., and you want to scan documents like that, this becomes a technique. Vector embeddings is a technique that can be applied to that same problem and help.”

Sisyphean development

Additionally, building out the data mesh does not stop the company from expanding, which means that the development is relatively continuous.

“It’s an ongoing process, because we‘ll have a new business opportunity, and it will require talking to a different accounting system,” Burmeister says. “I don’t know exactly how many accounting systems we have, but we have multiple, and we have not connected to all of them yet. I just heard about one the other day where it’s a different one that we haven’t actually looked at yet. So there’s an ongoing process to extend the coverage of the data mesh across SS&C, but it’s driven by client demand.”

Client demand prompted SS&C to build out an API capability to its data mesh that allows SS&C customers to access data easily from within the vendor’s private cloud. The private cloud allows SS&C to provision individual instances of the data mesh between the US, Canada, the UK, and the EU, so that data inside these instances is contained within those regions.

Burmeister says SS&C wants to build solutions on top of the data mesh in order to make it easier for application developers to build workflows and business logic on top of the mesh. By building out an extra step of publishing data queries as APIs, the team has made the mesh more accessible for developers, he says.

“We essentially turn on different capabilities based on what [clients] want, because we have hooked all the data together behind the scenes. As a result, we can also publish those APIs externally so the client can access that data,” Burmeister says. “So the client can say, ‘Hey, I want the actual accounting data that SS&C uses to generate my quarterly financial statements because I have some audit process that I want to do on that data.’ The client can call that same API, grab that same data and pull it over to their side and do whatever they want.”