Cost-Effective Edge AI Deployment: AI-Guided Logic Synthesis for Ultra-Low- Power TinyML Accelerators

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The growth of Edge AI and Tiny Machine Learning (TinyML) brings up a big problem between hardwareprice and processing power. Single Board Computers (SBCs) like the Raspberry Pi have enough computingpower for many tasks. However, their high price of ₱1,960 to ₱5,600+ and power draw of 2.5W to 15W makethem too expensive for battery powered IoT sensor networks. Microcontrollers like the ESP32 series aremuch cheaper at ₱110 to ₱280, but they have small SRAM and limited processing speed. This paper connectsElectronic Design Automation (EDA) logic tools with the cost of Edge AI. Based on the study by Astillero(2026), which uses Graph Neural Networks (GNNs) and Reinforcement Learning (RL) to reshape logiccircuits, we show that cutting down power and gate area at the circuit level helps run TinyML models oncheap sub-₱60 silicon chips. The results show that AI based logic pruning cuts dynamic power by up to 64%and gate area by up to 42%. This proves that AI assisted EDA makes Edge AI setups cheaper and morepractical for real world use.