This work aims to propose a bottom-up, two-step process that streamlines the design of analogue devices by using machine learning techniques. The complicated nature of these difficulties, which involve numerous variables and objectives, necessitates using designers' skills and knowledge while designing analogue complementary metal-oxide-semiconductor (CMOS) integrated circuits. The study offers a framework detailing the unique characteristics of creating analogue circuits using machine learning, and it looks into the potential of libraries that contain open machine-learning models to assist designers. Traditionally, commercial CMOS or software simulations have been used to create neural network designs; however, these methods may not always provide the best results. A three-stage device design is used to validate the suggested method. Using a machine learning technique called the decision tree; the stage type is correctly predicted with an accuracy of 89.74% in the first phase. To create prediction logic, two rule induction techniques are also used. In the second step, four learning techniques, decision trees, random forests, gradient-boosted trees, and support vector machines, are used to forecast the typical parameters for each stage type. The support vector machine yields the best results and has the lowest error rates of all these methods. © 2023 Elsevier B.V., All rights reserved.