Benefit from Expert's Know-How: Leverage the accumulated knowledge and insights of AI experts proficient in training autonomous agents using reinforcement learning. With Realmind, you gain access to tried-and-true methodologies that optimize learning outcomes.

Streamline with Delfox's RL-Ops Platform: Seamlessly automate your training infrastructure tasks through our state-of-the-art RL-Ops platform, 
collect and centralize training artefacts, ensuring you stay in control of your experiments and evaluations at all times.
Accelerate and reduce the complexity of RL projects with Delfox's proprietary coding framework: use RLFW simplified scaffolding of new projects, our custom models, simulation environment connectivity, training curricula extensions and more. Our framework also paves the way for maturation and explainability of trained agents, laying the foundation for real world integrations.

Elevate your projects with a solution designed to solve major pain-points of reinforcement learning applications.



1. Connect any simulator 1. Connect any simulator

The first step with Realmind is then to connect the simulator.
Benefit from our of-the-shelf package "Realmind Connector for Unity" to get started with your first training in minutes. Or rely on our expertise in connecting tens of simulators to get started with your project in Unreal Engine, Anylogic, Gazebo, and more.

Learn more
2. Code your own project 2. Code your own project

Realmind will generate a project structure which you will download to your preferred IDE using a classical Git connection.
RLFW - a Delfox framework for RL projects – will help you clearly structure definition of environments and experiments, use of standard or custom models and algorithms, define the training curricula, the analytics and metrics.

Learn more
3. Setting up your training scenario 3. Setting up your training scenario

Leverage your expertise and rely on the abstraction of the Realmind scenario editor to define your problem in the most precise way. The Realmind Scenario Editor allows you to change the composition of the autonomous team, randomize the position and number of obstacles, define the location of agents, and more. There are countless possibilities for setting up the learning scenario to suit your use-case.

Learn more
4. Configure the learning model 4. Configure the learning model

Customize the learning process with Realmind. Pick from our wide library of algorithms, which stays up-to-date with state-of-the art algorithms coming from the most renowned research institutions. Go further with our Neural Network Editor, which allows you to go beyond simple densely connected neural networks and lets you add attention blocks and convolution blocks.

Learn more
5. Build and Train 5. Build and Train

Realmind will build your project and distribute the training workload to predefined calculation clusters, will collect training artefacts and centralize them for you in an intuitive interface for experiment tracking.

Learn more
6. Evaluate the performance 
of your autonomous agents 6. Evaluate the performance 
of your autonomous agents

Once training begins, evaluate the model's performance on the task at hand. Visualize its progress on easy-to-use charts, as well as agent behavior in our 3D visualization player. This will ensure that your autonomous agents behave as you intended and are ready to overcome any new situation with confidence.

Learn more
7. Export your trained model to a physical vector 7. Export your trained model to a physical vector

With Realmind's export functionality, you can download the model in a standard and interoperable format.

Learn more

Realmind schema

Realmind brings clear structure to your RL projects, allows quick projects scaffolding based on Delfox’s RL framework, takes all the infrastructure and Ops workload away from ML Engineer and proposes a centralized artefacts exploration for experiment tracking.

Be focused on what matters to you – experiment and innovate, train your autonomous agent!

Realmind  schema