As the fundamental machinery orchestrating cellular functions, proteins influence the state of every cell profoundly. As cells exhibit significant variations from one to another, analyzing the proteome on a single-cell level is imperative to unravel their behavior and development. Beyond proteomic composition, the cellular and temporal context adds another layer of complexity to biological processes. Whereas the analysis of the genome and transcriptome of a single cell can be done readily on a commercial scale, this is far from reality for proteomics. Proteome analysis is challenging due to the vast chemical complexity, diverse number of proteins, and their concentration range. Moreover, unlike DNA, protein analysis lacks a method for amplification and therefore requires highly sensitive methods. In recent years innovative single-molecule approaches have been proposed but they still face many technical difficulties, especially sample complexity and throughput. This thesis introduces the concept and development of a mechanistically novel system, termed blinkognition. It employs spontaneously blinking fluorophores for the targeted characterization and analysis of single molecules, particularly peptides and proteins. The fluorescence blinking is sensitive to the chemical environment. We proposed that the signal modulation due to the presence of a molecule of interest could be leveraged for peptide identification. A data-driven approach was employed, where a machine learning model was trained on fluorescence intensity time traces from synthetically pure samples measured in single-molecule total internal reflection fluorescence microscopy. The resulting model could discern different peptide-dye conjugates of peptides with different sequences, phosphorylation, and epimerization patterns. Furthermore, we proposed to leverage model uncertainty to filter low-quality signals. Our latest work showed that the method can be extended to encapsulated, labeled proteins and even different sites on the same protein. Combined with an adapted labeling strategy, the method shows potential for protein profiling with many opportunities for further development even beyond proteins. Beyond the single-molecule level and toward the measurement of proteins in cells, this thesis describes the initial exploration for a small peptide tag that can bind a fluorogenic dye. Previous work has reported high-affinity binding of a peptide that contains two serine pairs by a fluorophore modified with two boronic acid moieties. The reported system features high background in cells, potentially due to natively present targets. We proposed combining fluorogenic rhodamine dye with boronic acids for peptide binding to lower this background fluorescence. We implemented yeast display of peptides with two serine pairs separated by two or eight randomized amino acids. Screening these yeast libraries for a peptide candidate that can bind and turn on the bisboronic probes remained unsuccessful in part hampered by low dye solubility. Nevertheless, with improved dyes and extended libraries this approach holds the potential to provide a small peptide for protein labeling. A small tag in combination with a fluorogenic small molecule could be applied to targets that do not tolerate large protein modifications such as fluorescent and self-labeling proteins while offering the photophysical advantages of a small molecule.