In medical settings, IV infusions deliver intravenous fluids or medications by allowing fluid to flow from an IV bag through tubing and a drip chamber, known as an administration set. In this talk, we will show how SAFEflow uses embedded vision and machine learning to observe fluid behavior directly and turn this process into a continuously monitored, closed-loop system. We will focus on how the product works: the imaging approach, the vision pipeline, the algorithms used to estimate flow rate and infused volume and the control methods that enable accurate real-time adjustment. We will also discuss the engineering challenges involved in making the system robust in clinical settings, including the need for high accuracy and reliability.

