APEIRON (Abstract Processing Environment for Intelligent Read-Out systems based on Neural networks) is a framework built upon a distributed computing general architecture for heterogeneous platforms and its software stack (from device drivers to high level programming environment). This framework is meant to be effectively used in the study, prototyping and deployment of TDAQ smart systems for HEP experiments. The architecture is organized in layers of distributed processing nodes, starting from a number of data sources (typically detectors or sub-detectors), following a streaming approach from data readout to low lever trigger or data storage. The computing platform consists of a modular and scalable network infrastructure, and a specialized set of computing devices (CPU, FPGA, GPU) in order to exploit hardware acceleration and use case adaptability in terms of latency and computing throughput.