Edge comprises a myriad of devices spanning micro-controller based, application processor based and server based. These are consumer electronic devices ranging from drones, small-robots, surveillance cameras, AR/VR goggles, wearables and many more. The compute type ranges from CPU, GPU, DSP, NPU and uC. These devices have different OSes such as custom-linux, ubuntu, android and ROS to name a few. The on-device acceleration libraries range from pytorch, tensorflow, tensorRT and such. What should an application developer do to enable a use-case? Too many choices, complexities in a compute, storage, memory and power limited Edge environment. We are building an on-device AI/ML library and API that lets application-developers build any conceivable AI/ML application with ease and also enables on-device training of few NN layers right on the Edge.