As global warming brings forth extreme precipitation in the typhoon-prone region in Asia, e.g. Hong Kong, Taiwan and Japan, the need for a scalable and yet accurate landslide sensing system becomes increasingly urgent. Natural slopes are the focus since their failures are often large-scale and disastrous; however, it is extremely expensive to monitor every single hazardous natural slope in existence with the current monitoring system. The MEMS (Micro-Electro-Mechanical Systems) technology which makes miniature yet accurate sensors available (e.g. accelerometer, gyroscope and magnetometer), will be pivotal towards the realization of a scalable landslide warning system. In this paper, we present the Smart Soil Particle (SSP) sensing system we build to detect the precursory tilts of a large-scale potential landslide site in LuShan, Taiwan. The SSP consists of the sensing module (i.e., low-noise MEMS accelerometer and a motion sensor) and the processing module (Raspberry Pi, a single-board computer). Both cost less than HKD$850; the price makes it possible to deploy a huge amount of SSPs on the slope surface without the fear of losing any. The processing module is capable of handling a large programmable data transfer rate (up to 800 samples/second/channel) through a 3G modem. In the back-end is a decentralized structured data storage system constructed to handle the massive data inflow from multiple SSPs. The whole system is built to send and receive data at a near real-time speed. With a proper noise control and shielding of the wires, we show that the low-noise SSPs are able to capture a small earthquake and also tiny vibrations due to cars passing by.