The Paradigm shift from Econometrics to Machine Learning

Our technology stack automates trading and quoting policies with the superior predictive characteristics of deep learning. We apply neural networks to the microstructures of auctions to exploit the maximum predictive value from all patterns inherent in the data feed and constantly adjust to new market regimes.

Since market making has gone under growing pressure from regulation and rapidly changing technology, continuous process automation, and risk minimization is rising the demand for artificially intelligent systems. Market makers can deploy AI systems to generate new quoting policies and creating risk monitoring systems for their existing models.

By viewing market making as an integrated process be optimized as a whole, deep-learning algorithms offer a promising approach for an entirely new market maker in emerging asset classes, such as Cryptos.

End-to-End Approach

Using deep learning for trading and quoting policies allows for an integrated approach with a completely differentiable algorithm, which can decrease the model risk caused by individual component optimization.

Unbiased Signal Generation

Policies generated by neural networks do not depend on human assumptions, but purely on patterns in the microstructure and update when regimes are changing.

Sophisticated Backtesting

Sophisticated backtesting and resimulation environments are the backbone of the development and improvement of AI-based policies. In addition, they allow for better model risk evaluation, more accurate P&L forecasts and the better understanding of historic losses.

Our Solution Scope

End-to-end automation of market making

AI policy systems for high-frequency products that minimize model risk without significant overhead. AI prediction systems for medium-frequency products which serve as an additional analytics tool for trade/quoting decisions.

Smart order execution and matching with deep learning technology

Order optimization layer for execution management systems. Predictive inventory management layer for in-house order matching.

Data Feed augmentation with predictions and backtesting environments

Data feed enrichment with deep learning based predictions, provision of complete (re-)simulation environments, liquidity monitoring, and detection of fraudulent behavior.

In-house cryptocurrency market making

We are providing liquidity to crypto & their derivative markets by using neural networks to generate quoting policies and predictively manage the inventory. Crypto markets are a valuable opportunity for us to build up distinct expertise in a growing asset and represents an environment where we test our latest technology.

Founding story


We started MX Trading ...

In early 2017, Merantix began to explore financial trading together with an industry expert who has a long track record in medium and high-frequency market making. For these frequencies, the amount of order book updates available make it an optimal use case for neural networks. Early on we used ML approaches for signal processing and deployed an LSTM model to achieve accurate time series predictions. MX Trading then began to live trade on established FX and crypto markets to gain relevant experience and establish the necessary trading infrastructure. After moving to market making, we received requests from specialized financial firms wanting to implement our DL models on their infrastructure. Seeing as MX Trading’s core competence is the application of deep learning technology to financial trading, we began to offer a range of solutions for financial institutions.
  • Luca Harrichhausen
    Luca Harrichhausen Entrepreneur in Residence

    Luca is an Entrepreneur-in-residence at Merantix. He graduated in Economics from the University of St. Gallen (HSG) and National University Singapore focusing on financial economics and monetary theory. Before joining Merantix, Luca gained experience as a Consultant at McKinsey & Company advising major exchanges and banks in Europe.

  • Filippo Scopel
    Filippo Scopel Machine Intelligence Engineer

    Filippo is a Machine Intelligence Engineer at Merantix. He holds a B.Sc. from the Technical University of Munich and a Masters Degree in Banking & Finance from the University of St. Gallen with a strong focus on Financial Econometrics and applying Machine Learning to Finance. Before joining Filippo automated portfolio constructions for Julius Bear and most recently gained deep experience in the application of Machine Learning to signal processing for a Munich-based IoT Company.

  • Dr. Jonas Probst
    Dr. Jonas Probst Machine Intelligence Engineer

    Jonas is a Machine Intelligence Engineer at Merantix. After graduating with a B.Sc. in Physics from TU Munich and with a Master’s in Mathematics from Cambridge, he conducted research on string theory for his PhD in Theoretical Physics at Oxford. He worked as a Quant at Barclays Investment Bank before joining Merantix.

Careers at Merantix

Engaging Topics

Merantix encourages you to pursue your curiosity and passion for several diverse topics. We are dedicated to addressing machine-learning problems that are challenging and foundational to an entire industry. We believe that diversity across topics is essential to attracting and retaining the greatest talent and to promoting creativity, continuous improvement, and functional learning.

Professional Development

Merantix is proactive in making major contributions to the open-source community. We encourage all of our employees to regularly attend professional conferences. Merantix is a regular guest at notable academic conferences on machine learning such as NIPS, ICCV, ICLR and ECCV. At the ECCV event this September, we will host an Italian Night with great food where you will be able to get in touch with all of us.

Work-Life Balance

Merantix offers a dynamic and flexible work environment. Some of us come in during “regular” business hours, and some prefer to start later or earlier. Our employees have the freedom to work when it best suits them. We are a fully trust-based organization that does not track vacation time. We believe in ownership of work, personal responsibility, and the motivation of our people, which has proven to be a successful approach to leadership


Collaborative Culture

Merantix is a family of diverse people from all over the world that came to Berlin to work on machine learning. We truly care about the individual well-being of our employees and support all of our new employees who are joining us in Berlin with whatever they may need. We have regular offsites and encourage everyone to organize common activities; just join the “all-sports” slack channel and stay up-to-date on bouldering, swimming, running, and much more.

Our experience