Jeremy Sosabowski of AlgoDynamix on using AI to forecast global markets
We focus on another finalist in the Cambridge Independent Entrepreneurial Science and Technology Awards
Jeremy Sosabowski is very good at telling you what will happen in the future.
In fact he or, more precisely,the company he co-founded, is certain of it.
What, then, does he see coming for AlgoDynamix?
“Growth, growth, growth, growth,” he replies. “We are very good at this one thing, so we should make the most of it.”
The calls keep coming for AlgoDynamix, a highly innovative risk analytics company with a base at the Cambridge Judge Business School and a business development team in London.
The firm is about to announce its latest big deal – with a major investment bank in the Netherlands. It’s also just been signed up by a large asset manager in Germany.
And a high-net worth individual in the United States has launched a fund for investors who want their money managed with the aid of AlgoDynamix’s artificial intelligence technology – a model that could be replicated in the UK.
The company uses machine learning to warn customers – typically banks, hedge funds and asset fund managers – about disruptive events in global markets. It anticipates price movements hours or even a couple of days in advance.
“If it’s asset management – pension funds – we can quite easily add an additional three to five per cent on their portfolio per year, which is humungous,” says Jeremy. “And it’s doing so in a very low risk fashion. They are taking much less risk to get these returns. It’s a double sugar coating.”
Lower risks, greater returns – no wonder Jeremy is forecasting growth.
But intriguingly, AlgoDynamix didn’t begin life as a fintech firm.
After his PhD in electronic engineering at the University of Cambridge, Jeremy span out a couple of companies. AlgoDynamix was, indirectly, the third and was co-founded with fellow experimental physics expert Wei Teo.
“We started this off in the medical space,” says Jeremy. “We were looking at cardiac arrests – can we see these coming? Would it be meaningful if we could tell you half an hour in advance?
“We made good progress, began clinical trials and had a good relationship with Addenbrooke’s.
“The idea was you would have a wearable device and it would give you half an hour warning of a heart attack by flashing.
“The best application was for the most vulnerable. If you’ve had a recent cardiac event, they can’t operate so you have a few months where they give you a defibrillator vest.
“It’s a horrible device and no-one wants to wear it all the time. So the idea was you had our device and when it flashes you put on the vest. Unfortunately, it didn’t work out commercially.
“It used very similar algorithms to what we use now. We didn’t have to reinvent the wheel – just find a different application for that wheel.”
Just as the technology was able to warn of impending disaster for patients, so it was able to detect disruptive events in financial markets.
“We closed a round of venture capital funding from Amadeus Capital Partners,” Jeremy recalls. “Our first major contract was a smaller financial institution in Singapore. They ran a large portfolio and had a lot of financial exposure so they said: ‘Jeremy, can you forecast the next big down – or up?’
“They were really keen and started paying us. We didn’t quite have the product ready then but we sensed there was a clear demand there – that’s how we pivoted from medical to financial services.”
The robust algorithms proved their versatility and were even deployed in the defence industry – but, predictably enough, Jeremy can’t say much about that.
“It used the same mathematics – the same engines, libraries and algorithms,” he says. “I think that’s why Amadeus was interested: potentially we had five or 10 different markets for it.”
AlgoDynamix’s risk analytics engine is based on ‘deep data’ agent-based algorithms. It scans primary data sources – the global financial markets – in real-time to analyse dynamic behaviour of buyers and sellers and forewarn customers about major ups and downs.
“In terms of alternative competitive solutions, traditionally even large investment banks have, ironically, always looked at yesterday to get insight for tomorrow. They try to infer from the past, which is interesting,” says Jeremy, wryly.
“We fundamentally disagree on that. If it’s business as usual, past data can probably help you. If you’re about to get something very big – the US election, the Brexit referendum, the French election results – you’ve never seen those events before. So we’ve taken a very different approach with our AI/machine learning. We don’t look at the past. We only care about what people are currently doing.
“We don’t look at news, historical data, analysts or anything except the data flow directly from the global financial exchanges. We’ve covered most of the exchanges now.
“By looking at the data flow and doing some careful analysis with our machine learning we can spot common behaviour. It doesn’t mean people are panicking yet, or the price is moving yet.
“But if everyone starts behaving in the same way and everyone has the same digital fingerprint, we spot that. Most of the time it’s random and you have no insight. When it’s not random, that’s when it gets interesting.
“It’s the herd mentality – when people start running for the exit. That’s what we’re doing with our clever algos: picking up the non-randomness.”
Spotting a trend before it becomes a trend is, Jeremy acknowledges, a challenge.
“It’s a tough problem mathematically and there is a lot of data,” he says. “The trend is not there yet. It doesn’t exist by the time we tell you. We will tell you 24 or 48 hours in advance. People don’t even know they are behaving the same way as their neighbours.”
The company says “noise classification, cluster identification and behavioural finance theory” are part of its core capabilities.
Jeremy prefers a vodka bar analogy.
“Imagine you have a big crowd of people at a networking event and there’s going to be a free vodka bar at 6pm,” he says. “They all start behaving in a certain way, clustering together and getting ready for it. They probably don’t know themselves what’s going on but it is happening and you can spot that if you’ve got the right type of mathematics.
“Tomorrow it might be very different. There will be a different crowd, a different networking event. It doesn’t matter. We don’t care what happened the day before.
“You will find a lot of predictive models out there. But what we’re doing is not predictive. It’s already there. It’s a bit like weather forecasting – the cloud formations are there but unless you have satellite imagery you can’t see it. Likewise with us, unless you have our clever algos, you can’t see it.
“So it’s never wrong because it’s telling you what’s happening. What can happen is for the crowd to break and diffuse but we’ll tell you that too. But most times, they go haywire.”
AlgoDynamix offers two products – the RAP platform to warn of major downside events, which particularly suits the insurance market, and the fast-growing ALDX PI platform for traders and investors, offering upside flags and ‘end of market direction’ flags, to help enhance portfolio performance.
“We see ourselves as augmenting the capabilities of portfolio managers and traders. We make them superhuman – dramatically more productive,” says Jeremy.
The PI platform is proving the most popular.
“It’s easier to sell the honeypot than doom and gloom,” observes Jeremy. “As of three weeks ago, in time for London Fintech Week, we enhanced the PI product even more adding more financial markets – Malaysia, Singapore, Indonesia and the BRICS countries – as well as the G10 countries, so we’ve been expanding big-time.”
And he revealed: “There is going to be another major update on the PI, probably early in quarter four. Customers are asking for it, so we’re building in the features. We’re being very pragmatic about the product roadmap – asking what people want, rather than what we think they want.
“As far as we know, nobody else is doing this. Some people might be doing it in-house but offering this level of advanced technology as a service to everybody is unheard of. It’s disruptive tech and a disruptive business model too.”
What happens though, if all the major investment banks sign up? Will the competitive edge be lost?
“It will be a nice problem to have!” suggests Jeremy. “We’ve had discussions with the big banks – ‘Three of your competitors have it and you’re the last not to have it…’
“But across our client base, they are all using it in very different ways. Just because we provide very similar insights doesn’t mean they will all react in the same way. They will respond based on their business and investment requirements.”
A report by EY recently suggested: “The biggest near-term threat to most banks comes not from fintechs but from traditional competitors better leveraging those fintechs.”
“Suddenly, they will have a massive competitive advantage and will move the industry. It’s happening already for us in Amsterdam,” he says. “It’s not us by ourselves but us partnering with these big banks. That’s the biggest change I see in the next six months.”
Longer-term, you have to wonder if we’ll even need humans to make investment decisions.
“Short- to medium-term, I think we’re augmenting. We still tell computers what to do. Rogue computers going haywire – I’m not buying it. I don’t see the Terminator thing happening…” he says.
He does, however, think further collapses like 2008’s are possible.
“It probably will happen again. Unfortunately, we don’t learn. We’ve had bubbles and crashes since the tulip craze in Holland in 1634. There are huge default rates on student loans in the US and UK. I don’t think it’s systemic so it might not spread but I definitely think it’s not being managed properly,” he suggests.
It’s not difficult, however, for AlgoDynamix’s charismatic CEO to forecast an upside event for the company.
“It’s been a good journey for us,” he says. “And we have very happy customers!”
• AlgoDynamix is among the finalists in the Cambridge Independent’s Entrepreneurial Science and Technology Awards. Read about them all in our Events and Awards channel.