Self-optimizing Customer Journeys

Award-winning Artificial Intelligence solutions based on adaptive Deep Reinforcement Learning.

Our specialty: Mobile Commerce.

Our partners



Today, optimizing the customer journey is done using simplistic statistical methods like A/B testing.
We provide a solution that handles personalization over the whole customer lifecycle out of the box.

Conventional approaches

Conventional approaches are iterative by design. Therefore, experiments can only adapt to environmental situations that lie in the past, and become outdated very fast.

Campaigns are usually optimized in isolation, without regards to the bigger impact on the customer lifecycle. While this simplifies the algorithmic challenge, it also severly limits the total achievable uplift.

Conventional methods have a significant need for manual intervention. They usually demand a constant creation of hypotheses, checking of segments for consistency and interfacing with IT for implementation.

Explored uplifts suffer from a significant delay between findings and their exploitation. In dynamic environments, that can even lead to vanishing uplifts.

Conventional Machine Learning is heavily dependent on the quality of a process called feature engineering. This approach is very expensive, error prone and not robust to changes in the environment.

Many classical Machines Learning algorithms are not easily scalable. That leads to severe problems as the data set grows.

Free Machines

By fundamental design, our algorithmic core is able to continuously adapt to new situations. Designed as an end-to-end decision system, it is built to be deeply integrated into production systems and therefore can instantly react to changes in the data.

Instead of focusing on single campaigns, our algorithm enables a sustainable optimization of the customer lifecycle. We implement custom target definitions in tight collaboration with you.

After the simulation period, where the system learns behavioural structures of your business and your customers, the systems constantly moves towards a very high degree of automation. With extensive monitoring capabilities, we allow you to focus on the big picture.

The system continuously learns and is able to exploit these findings in real-time. You'll never again will have to wait for the next optimization iteration.

free machines is built upon state-of-the-art Deep Learning approaches. Manual feature engineering is replaced by deep network architectures that learn the optimal feature set automatically.

Our algorithmic core is built on highly scalable tensor arithmetic, it is extremely scalable on GPUs. Data access is handled by a redundant, high-availablility web interface. Therefore, we are capable to handle very large datasets.

Use Cases

A general framework to increase long-term customer happiness and revenue.

B2C subscription businesses

Acting between data store and marketing tools, the Free Machines engine maximises the customer lifetime value by triggering the right E-Mails, SMS etc. at the right time.

Web shops

To dramatically reduce the amount of returned goods, Free Machines learns to identify customers with high-return behavior. Existing systems are then used to encourage multiple sales etc.

Online video services

In video ads, the Free Machines engine learns to strike the right balance between ad density and long-term customer satisfaction, depending on each individual situation of each customer in real-time.

Onboarding that minimizes risk

A gradual method to roll out a system that can grow in scale and scope.

1. Simulation

Based on existing data and current decision regimes, we build a simulation that provides our algorithm with a precise map of how your customer journeys work.

2. Trial implementation

With the simulation results as a starting point, we implement the decision engine on a growing portion of your user base.

3. Full scale implementation

After the initial installation is done and fine-tuning has led to proven equilibra, we scale the coverage of the system, targeting full-scale uplift.


Profound market knowledge.

Dr. Hannes Lüling


Hannes holds a PhD in computational neuroscience from the Technical University of Munich. After several years as Senior Data Scientist at ProSiebenSat.1 Digital and BMW, he is now heading AI research at Free Machines.

Dr. Olav Stetter


Olav is a physicist by training, with a PhD on the dynamics of complex systems. He also has extensive experience building technologies, teams and products in lean startup environments.

Dr. Uwe Stoll


Uwe has worked as a Senior Machine Learning Expert for several startups and enterprise clients. He can build on a PhD in Semantic Web and Machine Learning, during which he did extensive research in the field of e-commerce.

A radically simple pricing, ensuring positive ROI

We succeed if you succeed.

1. Simulation

Rebated daily rates
  • Onboarding
  • Data Science

2. Trial implementation

10% of uplift
  • Data handling fees based on complexity
  • Full daily rates

3. Full implementation

20% of uplift
  • Data handling fees based on complexity
  • Full daily rates


Work with us on the future of Artificial Intelligence.

Working Student: Real-time Web Development

Location: Munich, Germany

Job ad (PDF)

Freelancer/Working Student: Growth Marketing

Location: Munich, Germany

Job ad (PDF)

Where we are

We currently reside in the "Retailtech Hub" in Munich.

Contact Us

Feel free to leave your contact details, and we will get back to you as soon as possible.

Your message has successfully been sent.

Something went wrong!