Laya-MLX Per-Turn Classifier

What it does

M37 Labs demonstrates an encoder-only Laya-MLX architecture as a per-turn classifier for typed decisions.

How you can use it

Start by listing the questions you need answered from incoming customer messages. Note whether each question is a choice between specific teams, a scale, or a simple yes or no. You can also set a rule that sends the message to a human reviewer whenever the model feels unsure.

A developer can then set up the demo code to classify text quickly. This system only runs directly on newer Apple Mac computers with Apple chips. It will not work on standard Linux servers or older Intel Macs, and each message must stay within a short text limit.