Justin Lyon CEO of Simudyne talks 'simulation-as-a-service'

What is the future of simulation modelling?

Big Data and advanced analytics have been global tech trends for some time. But what about taking it a step further into the world of simulation? British startup Simudyne has a platform that facilitates the modelling of complex markets and claims to be a “step change in what can be done with agent based modelling, AI and simulations”. We caught up to with CEO Justin Lyon to learn more. The lightly edited Q&A can be found below.


Can you explain what your simulation service is?

Simudyne has built the next generation simulation platform, Providence, that is fit for global enterprises to test outcomes in a safe virtual world. Think simulations with billions of moving parts, millions of people interacting and the ability to decide the best course of action.


How does ‘simulation-as-a-service’ differ from the creation of digital twins?

The concept of digital twins concerns creating virtual representations of physical assets. While this is an interesting way to take advantage of the developments in simulation science and computational power, simulation-as-a-service is more general than that. Simulation-as-a-service can be applied to study social systems and, in particular, economies. Embedding artificial intelligence into agent based computational economic systems allows individuals within the simulations to learn from their interactions with each other and so more realistically represent the real world. This can help address a number of problems with traditional economic models of the world that assume people are rational maximising machines.


How did you get into the simulations market? 

I successfully exited two companies and following 9/11, decided to study system dynamics at MIT. I focused on simulating insurgencies and counter-insurgencies, ultimately supporting the war-fighters in Iraq as a contractor for the US Department of Defence. More recently, I worked on simulations at the Bank of England. I also got the chance to work with three major oil companies, two Banks and a massive insurance company. In every case, I got exposed to executive insights into extremely complex markets and situations, all only addressable with computational simulation.

Simulations of human intelligence, that is, Artificial Intelligence, will outstrip previous industrial and social transformations by using concepts and advances in computer science, robotics and mathematics to give machines the ability to think analytically and creatively. They will do it very differently than how we do it with our wetware, but they'll do it nonetheless. 


How big is the market for simulation and what is your differentiator?

The global advanced analytics [and] simulation market is [worth] $30-40billion. Providence has been built over the course of five years in collaboration with some of the biggest companies on the planet. The platform is robust and enterprise ready for global clients. It handles some of the most pressing concerns with simulation platforms: security, scalability, usability and flexibility. Providence is not a walled garden; all aspects of the platform are fully customisable with the skillsets you’ll find in any global enterprise.


How is this solution predominantly being used?

The Providence platform is used by modellers and developers at Banks and Financial Institutions to create AI-powered simulations for predictive analytics. They are used by the finance, risk and strategy functions to evaluate economic scenarios, regulations, risks, and strategic choices, with a focus on mortgages and stress testing.


Have any use cases surprised you so far?

We got the opportunity to help the CEO of a Bank with a number of subsidiaries in the Middle East successfully reach their ambitious growth targets whilst the countries in which they were operating dissolved around them. 


How might it be used differently in future?

There will come a time when all decisions of consequence are subject to a computer simulation. The injection of Artificial Intelligence into simulations will open up a Pandora’s Box of possibilities. The benefits will come on a micro scale, as we become able to simulate learning in real time in virtual worlds. They will also come on a macro level as we can create AIs that observe simulations of the real world and suggest interventions to achieve particular goals. For example, questions such as “How should we set interest rates in order to achieve price stability?” will seem like a trivial question when we can run simulations of the entire economy millions of times under all sorts of conditions and uncover the optimal set of policy tools. 


Can you explain how you are able to model whole markets – like the UK mortgage market – on the system? 

We create entire markets from scratch using individual specific data. This means data on each and every individual: their income, their age, number of children, current residence, etc. We then recreate virtual versions of each individual and inject them with algorithms that use this information to inform their behaviour. We then simulate the interaction of these individuals over time in response to various scenarios. For example, what happens to the unemployment rate if we have a recession and GDP falls by 1%? These models are so large and complex that we need hundreds or thousands of computers to execute them. We link these computers up with modern data processing technology and use their combined power to process the enormous amount of data and interactions. 


Your last round of seed funding was $410k in April last year – do you have more plans to raise capital? If so, what would you do with it?

Our [total] funding to date has been £1.1 million and we’re currently raising a Series A. We have ambitious plans.


Is there anything else you’d like to share?

Simudyne are currently working with Oxford University on the UK housing model. We have partnered with Cloudera and Microsoft and are currently on the Barclays Techstars accelerator in London. We fundamentally believe that all decisions of consequence should be subject to a computer simulation.