Machine Learning for Efficient Substitution Control toward Azido-Substituted l‑Sugar Synthesis via Flow Chemistry

Over functionalization of sugars under condition-dependent
constraints
without disrupting their native architecture remains a significant
challenge in vaccine development. Here, we report an AI-guided, automated
flow platform with variable reaction conditions that enables azide
incorporation at the C2 and C2–C4 positions of l-rhamnose
and l-fucose derivatives, achieving yields of up to 90–97%.
This approach delivers a safe handling of NaN3, minimum
human intervention, and approximately 3000-fold enhancement in space–time
yield compared to conventional batch synthesis. Subsequent in-flow
Cu-catalyzed azide–alkyne cycloaddition (CuAAC) affords mono-
and ditriazoles, offering a scalable route to glycoconjugates for
both medicinal and material applications.