Phenotype-based screening for drug discovery is increasingly employed in biomedical and pharmaceutical research. Recently, the zebrafish model has emerged as the main vertebrate model system due to its versatility and short life span. Thus, it would be very powerful model for high-throughput and high-content chemical screening experiments and large-scale phenotypic scoring. In particular, whole zebrafish in vivo approaches have the advantage that they can reveal toxicity and other side-effects of drugs at a very early stage of the study. Our engineering solution is to implement the in-line monitoring device like an office scanner combined with AI-based algorithm which enables sequentially observing large number of whole zebrafish as well as offering the quantitative phenotype information.