This thesis presents an automated imaging system developed to improve the monitoring of early life stages in fish, which are particularly sensitive to environmental stressors. Fish play a key role in ecosystems and food systems, but both wild and farmed populations face threats from climate change, pollution, disease, and overfishing. Monitoring young fish is crucial for understanding population trends and assessing the impact of environmental conditions, yet current methods are costly, time-consuming, and often involve sacrificing animals. The system described in this work replaces manual microscopy in laboratory studies with automated, non-invasive imaging, paired with software that uses machine learning and computer vision to extract key biological measurements. It enables the collection of time-series data with minimal human input, increasing data quantity and quality while reducing the need for lethal sampling. The system is validated through lab and field trials, including research vessels and tests with autonomous underwater vehicles, with results showing comparable accuracy to traditional methods but at a much higher efficiency.