Staying Ahead of the Game

We have developed an AI-based, patent-pending approach to help maintain long-term stable alpha contribution in investment products. Our proprietary, cutting-edge AI Platform automates the identification process of unique alpha sources.


With our AI Platform, we can identify unique alpha sources, often bespoke for specific requirements, in an industrial-like fashion. What usually takes months in development we can compress into mere days with our AI Platform.

This enables our clients to stay ahead of the game, fight alpha decay, improve their products, and protect their assets better.

Our application of Artificial Intelligence and Machine Learning

Although Artificial Intelligence (AI) and Machine Learning (ML) are often used interchangeably, they mean quite different things. AI is a broad catch-all term that describes the ability of a machine – usually a computer system – to act intelligently. In contrast, ML is the study of the algorithms and methods that enable computers to solve specific tasks without being explicitly instructed how and instead doing so by identifying persistent relevant patterns within the observed data.

We employ AI and ML for a vast number of tasks. Ultimately, AI/ML-identified alpha sources can benefit investors in two ways: diversification of portfolios, and reduction of risk.

Keeping in mind the investment management industry’s struggle to adapt to chaotic markets disrupted by topics like pandemics and wars, the case for investors looking to add non-correlated AI/ML-based alpha sources has never been stronger.

Client Benefits





What Differentiates our Approach to AI & ML?

We consider the use of AI and ML for the identification of alpha sources as complementary to the work of our capital market experts. Both approaches have their unique strengths. Domain knowledge experts can generalize their knowledge about one domain and transfer it to another. AI and ML methods can perform an exhaustive search over a huge number of potential patterns in the data and, in this way, find non-obvious, non-linear, multi-dimensional and fuzzy patterns that are hard to grasp by the human mind.

We benefit from both approaches not just by using them in parallel but by combining them into one synergetic approach:

Meet Dr. Dr. Roman Gorbunov – Head of Machine Learning at quantumrock

Going Beyond "Vanilla" Machine Learning

quantumrock´s Machine Learning methodologies have been adapted to our purposes, making our technology fully proprietary. We have identified a large number of proprietary methods to overcome the difficulties of working with financial data.


Financial markets are known to be very efficient, which leads to a very high noise-to-signal ratio in the financial data, which means that patterns/signals present in the data are obscured by considerable noise and, therefore, hard to detect.

Methods we apply to solve it:


The quality of an ML model is typically evaluated by comparing the model’s outcomes with so-called targets, which are present in the data. In the case of trading model development, the desired outcomes (optimal position/allocation) are not explicitly present in the data, making it impossible to use any regression model directly.

Methods we apply to solve it:


When building trading strategies, one needs to optimize financial metrics, like Sharpe ratio or total profit with slippage, rather than metrics typically used in 'vanilla' ML, like squared or absolute error. Therefore, one cannot simply apply standard ML methods to develop trading strategies.

Methods we apply to solve it:


In finance, one works with time series that closely resemble random walks most of the time. Standard ML methods are not designed to use this information and, therefore, are not suited well for the purpose.

Methods we apply to solve it:


In 'vanilla' ML, patterns are usually static (e.g., for decades, cats remain the same). In contrast, financial patterns are dynamic in nature, and they can evolve, disappear, and reappear. Therefore, one has to deal with a moving target, which needs to be appropriately addressed to produce strategies that can adapt to different states of the markets and, in this way, demonstrate a stable performance.

Methods we apply to solve it:

Automation of the Strategy Development Process

One of the ways to achieve a stable positive performance is to use a larger number of weakly correlated and properly weighted trading strategies, i.e. alpha sources. To practically implement this approach, we have developed an efficient way to generate new strategies that have an added value in the context of already existing strategies.

After being fed with data, our AI Platform generates fully-fledged trading strategies by automatically performing all the steps of the model development process (features construction, model training, testing, validation and statistical testing).

Human Input & Customizable Inductive Biases

We believe in a synergy between machine and human intelligence and, therefore, we combine a search for the patterns in the data with prior expectations provided by domain knowledge experts. Patterns found in the date get a higher significance, if they are expected or verified by financial experts. At the same time, some hypothesis proposed by experts can be rejected, if they are not supported empirically by the data. Even an automated data-driven search for a trading model is usually directed by a proper inductive bias embedded into the model architecture and constructed based on human input. By searching models in relevant and reasonable domains, we restrict the expressive power of models in a proper way and, in this way, significantly reduce the problem of overfit.