Laya the open source version of Jev

Laya is an open-source, non-autoregressive decision model that outputs calibrated probabilities for structured schema choices in 32.8 ms on a single GPU, outperforming Jev’s 150 ms latency. It uses bidirectional encoders and reinforcement learning to provide turn-by-turn conversion trajectories for vertical sales conversations, supporting over 100 languages with Apache 2.0 weights. The system replaces costly generative LLMs for reflex decisions such as ticket routing, spam detection, and urgency scoring.