Smart sensor nodes are key components of the intelligent systems enabled by the Internet of Things (IoT). The sensor nodes are unique in that they push the envelope in terms ultra-low power, form factor, and ultra-low cost, but these requirements cannot be addressed by conventional CMOS scaling. The sensor nodes are heterogeneous with building blocks spanning MCUs, non-volatile memory, RF, power management, integrated passives, MEMS resonators, and more. The power and performance of the sensor nodes are heavily dependent on the characteristics of the environment in which they operate. The design optimization of the sensor nodes involves hardware-software partitioning, analog-vs.-digital partitioning, and deciding between monolithic vs. package-level integration. In this talk, we discuss the role of system-level modelling in the design of always-on sensor nodes.