<![CDATA[Purpose – This study designs and evaluates an ESP32-based automatic ingredient dosing system for a small bakery experiencing imprecise manual weighing. Methods - A research and development approach produced two prototypes: an auger-based rice flour doser and a peristaltic-pump coconut milk doser. Both used load cells, HX711 modules, and single-speed on/off control with a 400 ms settling step. Manual and automatic dosing were compared over ten reported repetitions per condition using targets of 263 g flour and 492 mL coconut milk, the latter treated as 492 g under an assumed density of 1.0 g/mL. Findings - Reported accuracy increased from 91.5% to 97.9% for flour and from 95.2% to 98.9% for coconut milk. Mean absolute errors decreased to 5.4 and 5.3 g, respectively. Automatic dosing averaged 24.9 s for flour and 19.7 s for coconut milk. All nine functional test items were completed successfully. Research implications – The findings concern one recipe, enterprise, operator, and testing session. Manual and automatic timing covered different task boundaries, and calibration ended at 453 g, below the coconut milk target. Further calibration, equivalent timing comparisons, and longer-term testing are needed before generalizing performance. Originality - The prototype combines solid and liquid dosing mechanisms within a shared load-cell feedback framework intended for small-scale bakery production.]]>
