Bonsai

The Bonsai Artificial Intelligence platform provides developers, data scientists, and subject matter experts with the tools to facilitate the complete development lifecycle of an AI model:

• Build – The Bonsai web interface and command line tooling allow users to create AI models as well as develop programs for, manage, and utilize the AI engine.
• Teach – The Bonsai AI Engine and Inkling programming language enable developers to code, generate, and train AI models, independent of any specific backend algorithms or libraries.
• Use – The Bonsai generated API endpoints and SDK allow users to connect applications to, interact with, and analyze the AI models produced by the Bonsai AI Engine.

Bonsai brings together state-of-the-art techniques in machine teaching and machine learning, enabling enterprises to more effectively program and manage AI models.

AI-enable your development team – Bonsai allows developers to focus on programming concepts unique to a specific problem domain, leaving the management of complex, low-level AI mechanics to the Bonsai AI Engine.

Use existing simulation model – Using the Bonsai platform, enterprises can build a BRAIN (an AI model), connect the simulator of their choice, and train the BRAIN using that environment to learn a desired behavior.

Reuse and share your code and AI models – Programming of intelligence at a higher level of abstraction enables code and model reuse. System libraries and shared models can be leveraged across development teams.

Debug, inspect, and refine your AI – The high-level models produced by Bonsai enable you to understand what contributed to a prediction, identify conceptual gaps and bugs, and constantly refine your models.

Build models independent of the underlying algorithm – As machine learning and deep learning algorithms evolve, your Inkling code can be recompiled and retrained to take advantage of low-level technology advances.

The Bonsai platform provides an SDK for connecting simulations and modeling platforms for reinforcement learning and prediction. Bonsai utilizes MATLAB^® and Simulink^® as simulators for this reinforcement learning environment. Starting with an accurate Simulink system model and running thousands of simulations is critical in using deep reinforcement learning to determine optimal system parameters, such as tuning parameters for an industrial control system. The connectivity is either through a Simulink block, which performs a JSON-RPC call to the bonsai SDK, or through the MATLAB Engine API, where a MATLAB model can be driven in tandem with the brain models in bonsai.