The Runtime Formerly Known as TensorFlow Lite

LiteRT is the  Google on-device runtime for machine learning, formerly known as TensorFlow Lite. 

You can convert TensorFlow, PyTorch and JAX models to the TFLite format. 

This can be done using AI Edge conversion tools.

LiteRT rises to various ODML (On-Device Machine Learning) challenges:

1. Connectivity - ability to execute without an Internet connection

2. Size - reduced model and binary size

3. Privacy/data restrictions - no personal data leaves the device

4. Power consumption - efficient inference and a lack of network connections

Operationally, LiteRT models use an efficient portable format known as FlatBuffers, and the .tflite file extension. (See here for the difference between FlatBuffers and protobuf).

gRPC for .NET

The recommended library for using gRPC in .NET is grpc/grpc-dotnet under the Apache 2.0 license (The original implementation is now deprecated). 

gRPC is a (language agnostic) remote procedure call framework designed for high performance.

gRPC is often used with protobuf with coding structured around .proto files explained here.  One of the reasons for high performance is compact binary serialization of these so-called protocol buffers.

Common use cases include gRPC authentication (gRPC can converse with a variety of authentication systems), compression and other distributed communication scenarios. 

It's first commit was made internal to Google in December 2013 and was made public on Github in 2015. Google was doing RPC with protocol buffers for years and years,  but needed to expose APIs for public consumption. In 2017, gRPC was donated to the CNCF.  More complex features were incorporated like Flatbuffer, which is for serialization - where you can use the data without full deserialization (originally used for games). Observability features have also been built into gRPC since its initial release.

gRPC is used by Netflix and Facebook and Google of course.