Aminoacyl-tRNA synthetase (aaRS) engineering has enabled the co-translational incorporation of several hundreds of non-canonical alpha-amino acids (ncAAs) into proteins in living cells. These sequence-defined biopolymers hold immense potential in therapeutics by facilitating the synthesis of macrocyclic peptides, natural products, and antibodies in cells with enhanced chemical diversity and the possibility for evolution towards newto-nature biological functions. Moreover, its synthesis in the context of cells enables subjection to directed evolution for new-to-nature functions. While majority of the current repertoire of ncAAs have a large sidechain diversity, they have very limited backbone diversity (mostly alpha-amines), there has been a significant effort in expanding the backbone diversity of proteins by facilitating incorporation of alpha-subsituted monomers to generate sequence-defined biopolymers. However, one of the major bottlenecks is the scarcity of aaRS capable of acylating tRNAs with these alpha-subsituted monomers. Traditional approaches to evolve novel aaRS mutants typically rely on coupling the activity of aaRS to the expression of a reporter protein with a selectable phenotype (either antibiotic resistance or fluorescence). However, such translation-dependent evolution schemes are incompatible with monomers that are suboptimal substrates for the endogenous translational machinery (EF-Tu or ribosome). To enable the ribosomal incorporation of such exotic monomers, a two-step solution is needed: A) Engineering aaRS to acylate its cognate tRNA with these monomers, without relying on ribosomal translation as a readout, and B) Subsequent engineering of the translational machinery to accept the resulting acylated tRNA for translation. Here, we report a novel platform for aaRS evolution that directly selects for tRNA-acylation without ribosomal translation (START). In START, each distinct aaRS mutant is linked to a cognate tRNA containing a unique barcode sequence. Acylation by an active aaRS mutant protects the associated barcode-containing tRNAs from an oxidative treatment designed to damage the 3’-terminus of the non-acylated tRNAs. Sequencing of these surviving barcode-containing tRNAs is then used to reveal the identity of the said active aaRS mutant. The efficacy of START was demonstrated by identifying novel mutants of the M. alvus pyrrolysyl-tRNA synthetase from a naïve library that charge ncAAs. Moreover, we demonstrate that START can be used to detect aaRS acylation activity for monomers that are poor substrates for the endogenous translation machinery.