Exactpro is a testing business that specialises in complex financial infrastructures, in particular with the application of AI amd model-based testing. Its clients have included a number of leading exchanges, including the London Stock Exchange Group (LSEG), Nasdaq, Japan Exchange Group, Qatar and Athens, as well as clearing platforms such as ISX Financial, in Cyprus.
The Exactpro business was developed as an independent services provider under the ownership of the London Stock Exchange from 2015, and was bought out by its founders in 2018.
One of those founders is Exactpro’s CEO, Iosif Itkin. QA Financial met him to talk about the opportunity in AI and about his future plans for growth. But we started with the major reorganisation of Exactpro’s business that was forced on a still-young business following Russia’s invasion of the Ukraine in February 2022.
QA Financial: Many of your staff were based in Ukraine and also in Russia at the time of the invasion. You’ve had to move them and you personally relocated. Where are you now?
Iosif Itkin: Georgia is now our largest staff location and when I’m not travelling across client locations I spend my time in Georgia. We have more than 300 staff here out of a total of 500. The second largest location for us now is in Sri Lanka. And then we have people distributed across the globe: in the UK, in the US, Lithuania, Armenia, Serbia, et cetera. We completed the transformation of our business before the end of 2022, and what is most important is that none of our clients were affected or disrupted by the fact that we have to transform the company.
QA Financial: Exactpro specialises in working with financial markets customers. Is that still the sole focus of your business?
Iosif Itkin: It has certainly been important for us since we started. By focusing on financial markets infrastructures we were able to build a very strong expertise in a relatively narrow domain of exchanges and clearing houses.
And so far, most of the applications of our expertise have remained in financial services, because of the complexity of the customer requirements – for example, latency requirements and availability requirements – and because of the compliance requirements. Now we are increasingly using data-based methods, and these are more agnostic to the subject domain.
“As long as you throw enough data at the problem it is possible to apply the same tools and methods to different subject domains.”
– Iosif Itkin
The way of thinking about different industry domains is changing with AI and the growing importance of data. The same approach that we have for data processing for testing should be applicable to other domains. We have recently completed some testing work for ARTEX, a new Liechtenstein mulit-latereral trading facility that enables the fractionalisation of art. It’s an example of how our work can extend into other verticals.
QA Financial: If you were to diversify into other markets, would you need to be part of a larger organisation offering a wider range of digital transformation services – not just testing?
Iosif Itkin: We still believe that software testing is an important, and specialist, service. That’s not something just related to companies and how they are organised – it’s all about people. If you read blog posts by software testers, they are always trying to explain that they are not just doing testing. My take is that with the growing complexity of systems, providing information about technology is a very important endeavor.
But, going back to our plans for Exactpro, we do believe that one possible growth path is building joint centers of excellence with our key clients. Many customers would like to maintain strong internal expert expertise and capabilities. And we believe there are good business cases for building centers of excellence, jointly operated by our team and the client so that both can contribute their strongest capabilities. And also, testing is highly dependent on strong software development skills, because of the technical complexity that testing frameworks have to simulate and verify.
QA Financial: The idea of centers of excellence was fashionable; then it went out of fashion and now it’s coming back. Is that driven by the requirement of large-scale Cloud transformations?
Iosif Itkin: No, I would not say it’s specifically related to the Cloud. Intensive data processing for testing can be managed with an on-prem infrastructure, within the client’s data centers. For tasks that require only short bursts of computational capacity, Cloud can be a good option. But if you have many applications that are deployed and maintained daily – continuous delivery for a large set of systems – then it may be that the bill for the Cloud becomes enormous. There is still a strong case for using the client’s data center to build the data pipeline.
That’s part of the argument for the center-of-excellence model; we would be able to share your expertise among customers about how to connect data for testing to those. And a key point here is that the majority interest in the center of excellence should rest with the client, so they can be more confident in their regulatory obligations and their business continuity.
QA Financial: The key test automation tool Exactpro offers is called th2.What makes it different to rival test automation platforms?
Iosif Itkin: I believe that over the past two years we have acquired a deeper understanding of what is required for intensive data processing. And to a degree we have gone down the road of simplifying everything: machine learning means you have less code and more data. So test platforms should be as simple and as nimble as possible and our recent advances have come not from building more complex or sophisticated test platforms, but instead trying to simplify the test frameworks and tools that we use.
The platform that we are using across most of our new projects, th2-shark, is in fact a simplified version of our original th2 platform. It allows us to deploy the tool almost everywhere. And the other part of the solution is we can collect as much data as possible for the processing.
A key challenge for financial firms in applying AI to testing is to make sure that you have enough data in a normalized form. For example, encounter cases where we are asked by a customer: “This is our generator and we’ve produced 300 test cases. Is it enough for us to apply AI?” The answer is that this may be enough for some symbolic AI scenarios for testing. But if we are trying to use machine learning for subsymbolic AI, the volume and diversity of data that we need to use is on a different scale.
“Test platforms should be as simple and as nimble as possible.”
– Iosif Itkin
If a company is only able to collect data points that are numbered in hundreds, it is not possible to use AI in testing unless, of course, you are using some other dataset that is outside of the organization, such as ChatGPT. But that is not your data. So the first challenge is to collect enough data; submitting diverse inputs to your system and collecting all the outputs. That is not a trivial problem and many firms have not yet solved this first step. There may, for example, be very specific limits to the number of transaction data points you can collect from the firm’s UI frontends.
QA Financial: What is the main challenge financial exchanges face in terms of testing?
Iosif Itkin: It is complexity. Our customers are trying to deliver increasingly complex systems at faster times-to-market – which means information has to be extracted in a shorter period of time. So most of our customers are either undertaking major transformation projects or they have an inside schedule for a particular project. Most have their own delivery teams of course, but there is still plenty of room for co-operation with an external partner or in partner joining in a centre of excellence.
There are three main challenges facing organizations trying to apply advanced model-based testing. The first thing is that when you’re trying to build sufficiently complex models you cannot risk delaying the project while you’re building those models. You need to be able to run a testing twin.
Secondly, there is an art to model-based testing, which is to find what simplifications you can make while ensuring the model remains useful. And the third challenge is the caliber of people you need. Ideally you need people who are at the confluence of software testing, software development and data analysis. These people are not easy to find or to train. These are reasons our customers seek partnerships with Exactpro.
UPCOMING QA FINANCIAL EVENTS
READ MORE
- Kobiton launches new capabilities for mobile testing
- ECB names new digital chief as QA climbs the priority ladder
- Maveric eyes IPO amid major US growth plans
- Exclusive: ‘AI productivity means more risk,’ warns Tricentis ML head
- QualityKiosk wins major testing deal with Dubai’s largest bank
Become a QA Financial subscriber – for FREE
News and interviews * Receive our weekly newsletter * Get priority invitations to our Forum events
