Mass spectrometry is the cornerstone technology for large-scale proteomics, enabling the systematic characterization of proteins in complex biological systems. In a standard bottom-up workflow, proteins are enzymatically digested into peptides, ionized, and fragmented to generate tandem mass spectra. The central computational task is to map these spectra back to peptide sequences, traditionally achieved by database search engines comparing experimental and theoretical spectra. Recent advances in machine learning (ML) have transformed this process, improving sensitivity, speed, and confidence in peptide identification. As proteomics underpins modern drug discovery - by uncovering disease mechanisms, identifying biomarkers, and characterizing therapeutic targets - we will discuss key achievements of ML-boosted computational proteomics, current limitations, and the emerging challenges that must be addressed to accelerate therapeutic development.