In the modern world, cancer is one of the leading causes of death, and its early diagnostics remains one of the big challenges. Since cancer starts as a malfunction on the cellular level, the diagnostic techniques have to deal with single cells. Detection of circulating tumor cells (CTCs), which are present in the patient's blood, holds promise for the future theranostic applications, as CTCs represent the actual state of the primary tumor. Raman spectroscopy is a label-free technique capable of non-destructive and chemically-specific characterization of individual cells. In contrast to marker-based methods, the CTCs detected by Raman can be reused for more specific single-cells biochemical analysis methods. This thesis focuses on technological developments for Raman-based CTC identification, and encompasses the whole chain of involved methods and processes, including instrumentation and microfluidic cell handling, automation of spectra acquisition and storage, and chemometric data analysis. It starts with a design of custom application-specific instruments that we used to evaluate and optimize different experimental parameters. A major part is software development for automated acquisition and organized storage of spectral data in a database. With the automated measurement systems and the database in place, we were able to collect about 40.000 Raman spectra of more than 15 incubated cancer cell lines, healthy donor leukocytes, as well as samples originating from clinical patients. Additionally, the thesis gives an overview of data analysis methods and provides an insight into the underlying trends of the dataset. Although the cell identification models could not reliably differentiate between individual cancer cell lines, they were able to recognize tumor cells among healthy leukocytes with prediction accuracy of more than 95%. This work demonstrated an increase in the throughput of Raman-based CTC detection, and provides a basis for its clinical translation.