Bistable switches are ubiquitous in natural and engineered biological systems, playing a key role in cellular regulation. We constructed a bistable switch consisting of tetracycline operon, arabinose operon, and lactose operon, allowing for the artificial control of the rate at which cells absorb metabolites from their environment and direct the flow of carbon within the cell. This aims to achieve rapid growth of engineered bacteria and efficient production of target metabolic products. We employed error-prone PCR and machine learning-assisted directed evolution to optimize the bistable system, increasing its tunability. Mutants were analyzed by observing fluorescence intensity and using visualization tools such as Pymol and molecular docking, to identify those that bind tightly to DNA and loosely to inhibitors. Eventually, we individually screened out an araC and tetR mutant that theoretically bind tighter with DNA and show better inhibition effect compared with the wild type.