It's not what's next, it's what I'm doing now:
I am currently creating an API app with websockets and; a persistent subscription to any MQTT broker like the one I just released.
I am streaming real-time sensor (or otherwise) data into neural networks YOU create. The app integrates BrainJS, GraphQL, MongoDB, Redis, React, Docker and more...
I am targeting Q-learning, and I will be using Tasmota sensors (Sample Client) and commands.
- I am providing endpoints to permit real-time, zero-trust and concurrent data streaming I/O.
- I will integrate a previously created proof-of-concept BrainJS implementation to train and respond to neural network client requests that submit reinforced training data.
- I will standardize the UI using React (a dumb terminal), using Google Auth to CRUD your Neural Networks, Model Samples, Sample Clients.
- Using zero-trust API keygen on each neural network so zero-trust can be cached with Sample Client's streaming I/O training data.
- To provide intelligence.
The only way to answer if this has been done, is to do it.