VELMA: Crypto-assets discovery tool

Why use VELMA?

I’m into crypto. I have a nice portfolio, but I want to expand it. I scroll through my Twitter feed and the only thing I see are people trying to pump projects that make their life better. It’s difficult to find projects that fit my interests and investment strategy. I need a tool that lets me explore the crypto-asset space, tailored to me. Luckily, there’s VELMA, which does exactly that.

Crypto exchanges, wallet providers, portfolio tracking apps, media outlets… They can all benefit from VELMA to bring value to their end-users; leading to higher retention and profits.

In this article, we’ll tell you the origination story of VELMA, how it works and how your business can benefit from it. If you’re not interested in the inner workings, you can jump straight to the section: How VELMA brings value to wallet providers, exchanges, portfolio tracking apps, etc.

Soon we will launch a closed alpha. Interested in joining? Read more at the end of this post.

VELMA - How it all started: Hyper-personalization in wealth management

Hyper-personalization is present in every aspect of our lives; Netflix recommending series to watch, Amazon recommending books to read, Facebook ‘recommending’ ads to click on, and Youtube taking you down the rabbit hole. Those are the examples you read in every write-up on recommender systems.

At IntellectEU, we took the recommendation game to a different industry. An industry that is known for its manual work and in-person relationships: the wealth management industry. More specifically, we focus on the sub-industry of financial advice. At its core, financial advice is simple: advisors maintain and (try to) grow the wealth of high net worth clients. They have regular touchpoints with their clients in which they propose assets to buy and sell. They keep track of client data, product data, and portfolio data in TAMPs (turnkey asset management platforms). Yet, it still takes significant effort and research to come up with suitable recommendations.

When reading the last paragraph, the machine-learning-savvy readers might see this as a perfect environment for a recommendation engine. That’s why we’ve built VELMA, the machine learning tool that adds a layer of intelligence to wealth management platforms. VELMA’s upselling and advisory modules are currently being used by large wealth management platform companies to spot new opportunities at scale. VELMA does not replace human capital, nor technological systems and processes. It enhances them and frees up time to focus on what truly matters… Driving revenue and enhancing client relationships.

Recommending crypto assets

We’ve built VELMA in such a way that it can cater to any personalization/recommendation use case in any industry. We automatically build a dynamic knowledge graph based on input data.

Let’s now zoom in on the crypto asset recommendations. For this test, we got data from Messari. We included both descriptive data (token usage, consensus protocol, launch style, tags, the total developer commits, Reddit community members…), and performance data (volatility, ROI, transaction volume…). In total, we have 31 data fields.

Important to note is that we just need to drag and drop the CSV file and VELMA constructs the knowledge graph. As you can see, a lot of assets also have missing fields. VELMA can handle this without any problem. Here’s the expanded node view for one asset. Let’s take the all-time classic, Bitcoin, as an example:

In the next step, Velma translates the knowledge graph into profiles; a numerical representation of the crypto assets. This is where Velma innovates, we’ve built an unsupervised machine learning model that takes the knowledge graph as input and produces profiles.

Profiling is at the core of our product and fuels the other use-cases. On a high level; it is the creation of a profile for every asset based on all available data. A profile can be thought of as a set of numbers. If two assets are similar, their profiles will also be close together.

Let’s start with a very simple example to illustrate the concept of profiles. Imagine that we have several crypto assets. One way of comparing them is by looking at their volatility. Another way would be to look at their consensus mechanism. You could also make a comparison based on both the volatility and consensus mechanism. If you add even more dimensions, those kinds of comparisons quickly become very complex. This is where profiling comes in. Based on all the data you have on the assets you’re looking at; VELMA creates numerical profiles that can be easily compared to each other. The effect of the different kinds of information (e.g. volatility, consensus mechanism, developer activity…) on the final profiles is fully adjustable. Here’s a visual representation of the profiles in which we only took descriptive information (so no performance data) of the assets into account. As an example, you see the profile of Compound and its neighboring assets (which are indeed other DeFi protocols (Uniswap, 1inch, Curve, …)).

Once the profiles have been established, they can be used to power recommendations. In its simplest form, you could look at which assets your user already has, find the assets with the most similar profiles, and recommend those. If you also include user metadata (e.g. age, gender, capital to invest, preferences,…) in the knowledge graph, and crypto-asset portfolio data for those users; the recommendations get even more precise. But let’s keep that for another blog post!

How VELMA brings value to wallet providers, exchanges, portfolio tracking apps, etc.

How would all of this bring value to your business? Let’s say you’re a wallet provider or an exchange: Your users hold crypto assets, holdings that are shown in your platform. For now, they add new assets based on their own research or tokens being pushed on Twitter. Using VELMA, you can give your users a very intuitive way of discovering new assets, relevant to them. When those recommendations trigger their interest and they confirm the trade, you earn a transaction fee. Let’s say you’re a crypto research blog or media outlet: Your readers love your articles. Your articles have several tags; one of them is the crypto-asset that is covered in the article. By using VELMA, you can very easily recommend other articles that would be of interest to your readers; increasing time spent on your website; increasing ad revenue.

Let’s look at a few examples of how this works, illustrated by using VELMA’s frontend.

Important to note is that all functionality is provided by our API and integrates seamlessly in any back-end infrastructure.

In the above GIF, we show the recommendations for several inputs. For bitcoin, VELMA recommends other payment network solutions. For True USDC, VELMA recommends other stablecoins. For SushiSwap, VELMA recommends other DeFi projects, and for Algorand VELMA recommends other smart contract platforms. All of this, out of the box, leveraging our AI-generated profiles.

The above demonstrates only a fraction of VELMA’s capabilities. You can add filters on top of your recommendations, you can take into account the user’s token holdings to make the recommendations even more personalized, or you could look for the least similar assets to diversify a portfolio. Furthermore, you can also use VELMA’s prospecting capabilities in which you enter information about your ideal asset, VELMA builds a profile based on this recommendation and fetches the most similar existing assets as a recommendation.

Interested in learning more?

We will start a closed alpha phase in early November. For this closed alpha, we’re looking for a wallet provider, a media outlet, and an exchange to participate. Get in touch via velma@intellecteu.com to learn more.