DemaTrading: Making cryptocurrency investing accessible with AI and machine learning
About DemaTrading
DemaTrading is on a mission to make cryptocurrency investing more accessible, by reducing the risks and removing the need for expertise. Their plug & play solution for exchanges makes it easy to offer index funds to the exchanges' customers. Launched in 2021, it now has over 18,000 end-users, manages over €7m in assets, and handles over €250m in trading volume.
Tell us your challenge. We're here to help.
Contact usA rapidly scaling new company, DemaTrading was being held back by its existing architecture. Moving to Google Cloud gave it the infrastructure and machine learning tools it needed to scale up at pace.
Google Cloud results
- Only paying for resources as needed reduces infrastructure overheads to 80%
- Maintenance time is down from 4 hours a week to 15-30 minutes, allowing more focus on core priorities
- Downtime reduced from 5 hours a month to zero, improving the user experience
- Machine learning tools enable new customized trading products for the end user
DemaTrading scales up at speed with Google Cloud
The popularity of cryptocurrencies has grown at an astounding rate since its conception in 2008. Born against the backdrop of the global financial crisis, its decentralized, digital ledger system initially appealed to people with a mistrust of traditional financial institutions, but it quickly became mainstream, with 39.29 million cryptocurrency users in 2017 rising to as many as 257.2 million users in 2022.
However, with the collapse of the Luna crypto network, followed closely by the FTX crash, some of the risks of crypto investing have been brought into sharp relief, particularly for the inexperienced investor.
Alive to this issue back in 2021, Demian Voorhagen and Marijn van Rijswijck founded DemaTrading with the goal of mitigating much of that risk, in order to make crypto investing accessible to everyone, regardless of their background or experience level.
"A lot of people want to dabble in crypto," explains Demian, "but it's very complicated, which leads them to only invest in one or two currencies, giving them a very high risk profile. Or else they're not well informed, which means they invest in poor projects. What we do is give people an easy way to make safer investments in the crypto space."
Looking to the traditional markets, Demian and Marijn realized that indexes provided an ideal solution to the problem, and they set about building DemaTrading to offer index-based investment portfolios through cryptocurrency exchanges, allowing individual users to automatically spread their investments across an entire market or market segment. This not only removed their need for expertise, but also significantly reduced their exposure to risk.
Searching for the architecture to scale at speed
From its first pilot algorithm to offering a wide range of customizable portfolios to exchanges, banks and pension funds, DemaTrading has had to scale up rapidly in a very short timeframe. It quickly found that its existing infrastructure was not able to scale with it.
"As soon as we started increasing the complexity, the ease of use started to hinder us," Marijn explains. "We had fewer customization options, fewer ways of going about things, which really limited our progress. We spent a lot of time debugging and bug fixing, which ultimately hindered the development of new features."
"Google is famously known for its very sophisticated machine learning infrastructure, so that is one of the things that really appealed to us: the fact that we could easily scale up to a point where we wouldn't be held back by the features of our cloud provider. Wherever we want to go, our cloud provider should be able to get us there."
—Marijn van Rijswijck, Founder - Tech, DemaTradingWith its eyes on the future, DemaTrading decided to change its cloud infrastructure. Having already identified machine learning as a key component of its growth, it turned to Google Cloud.
"Google is famously known for its very sophisticated machine learning infrastructure," says Marijn, "so that is one of the things that really appealed to us: the fact that we could easily scale up to a point where we wouldn't be held back by the features of our cloud provider. Wherever we want to go, our cloud provider should be able to get us there."
The migration of DemaTrading's entire web stack took one engineer two months to complete, "which says something about the level of guidance that you get from the online resources," says Marijn, "as well as the ease of use of Google architecture. We were very pleased with it."
Improving efficiency with increased automation
With over 18,000 account holders, demand for the DemaTrading app fluctuates, as users log in to view their dashboards and make trades. However, DemaTrading's previous infrastructure used the same level of resources regardless of the level of demand, which meant that when demand was low, it was paying for resources it didn't need.
Now, DemaTrading's Google Cloud architecture scales up and down according to the level of activity, which means DemaTrading only pays for the resources it uses. As a result, overall costs have come down 80%.
DemaTrading has also noticed an improvement in its product's usability thanks to Cloud Run and Cloud Tasks. With its previous architecture unable to handle asynchronous tasks, users who had logged on to make a trade were forced to wait while the app processed their requests, which involved a lot of manual checking and button pushing on the developer side.
"Now, we simply write tasks off and put them in a queue and Cloud Tasks automatically picks them up and executes them, one by one, just like a worker. The user doesn't need to wait for those things to finish, which really improves their experience."
—Marijn van Rijswijck, Founder - Tech, DemaTrading"Now we simply write tasks off and put them in a queue," says Marijn, "and Cloud Tasks automatically picks them up and executes them, one by one, just like a worker. The user doesn't need to wait for those things to finish, which really improves their experience."
DemaTrading's lead engineer now spends around 15-30 minutes a week on maintenance and checking, down from 4 hours a week with the previous infrastructure. These time savings, along with the reduction in monthly overheads, mean DemaTrading has been able to develop a far more robust development environment, which in turn has resulted in a 20% improvement in the ratio of bugs to new features. The rollout of DemaTrading's open beta of its dashboard, for example, was made possible due to its confidence that it would have the ability to handle any problems if they arose.
As for DemaTrading's users, "they are experiencing far more stability and consistency in the rollout of new features," explains Marijn. "Before, our infrastructure wasn't capable of doing all these sophisticated things in a rolling way. So our users experienced less accessibility and uptime. Now, we have gone from having an average of 5 hours a month of downtime to zero, as well as fewer bugs, all of which makes for a far smoother user experience."
Personalizing product features with machine learning
As DemaTrading grows, machine learning is playing an increasingly important role in shaping its product offerings, which means it is important that DemaTrading is able to store and process its data efficiently.
"Previously, we were renting a big, chunky cloud instance and running our machine learning models there," Marijn explains, "but it wasn't very cost effective. It did the job in the beginning, but if you really want to scale it and make it cost efficient, then you need something dedicated for this purpose. And that is where Google comes in."
Working efficiently with large datasets in BigQuery, DemaTrading is starting to use machine learning tools to optimize on price execution. By training its models to trade during optimal time windows, rather than at one set time, DemaTrading can ensure that customers' holdings are bought and sold at the best price. Not only will this earn investors more profit, but by spreading trades out through the day, it will also give the exchange more liquidity, making it easier for other customers to buy and sell on the platform.
Another feature in development is a recommendation system using Recommendations AI to analyze end users' behavior and tailor specific products to their individual investment profile. If an investor is more risk averse, they will be able to select a highly personalized product that reduces their exposure to risk.
A future without limitations
For DemaTrading, these AI-powered features are all part of making crypto more accessible and less risky for the everyday investor. As Marijn and Demian plan for the future, Marijn is confident that DemaTrading has the infrastructure and machine learning capabilities to enable the company to grow.
"We just know that for 99% of our use cases, there is a feature and a guide to get us there, which greatly improves our cost efficiency and the swiftness with which we can deploy these new features, and ultimately keep the development train rolling."
—Marijn van Rijswijck, Founder - Tech, DemaTrading"The main thing is that we're not held back by our architecture anymore," Marijn explains. "Knowing that you have the tools at hand to achieve what you have in mind allows you to think more broadly and clearly about the value of the product you're developing, both in the short and the long term. We just know that for 99% of our use cases, there is a feature and a guide to get us there, which greatly improves our cost efficiency and the swiftness with which we can deploy these new features, and ultimately keep the development train rolling."
Tell us your challenge. We're here to help.
Contact usAbout DemaTrading
DemaTrading is on a mission to make cryptocurrency investing more accessible, by reducing the risks and removing the need for expertise. Their plug & play solution for exchanges makes it easy to offer index funds to the exchanges' customers. Launched in 2021, it now has over 18,000 end-users, manages over €7m in assets, and handles over €250m in trading volume.