We are developing a microfluidic droplet sorting platform to study co-cultures of yeasts and cyanobacteria. Currently, our system supports droplet detection, synchronized image acquisition, and sorting into specific output channels. The platform is capable of sorting droplets into three or more separate outputs, making it especially well-suited for co-culture research. One critical gap in our system is the absence of image-based analysis for classification and decision-making. To address this, we are utilizing a Raspberry Pi 5 paired with global shutter camera. To meet the real-time processing requirements, we plan to integrate the Hailo-8 AI Accelerator (available as an expansion board for the Raspberry Pi) and evaluate the complexity of neural network models to achieve our throughput goal of 100 droplets per second. Alvis computational resources will be used to train various neural network models for tasks such as image segmentation, object detection, cell classification, and general image processing. Once trained, these models will be compiled and optimized for the Hailo-8 AI Accelerator, enabling real-time image analysis and droplet sorting on our platform.