In this research, an investigation is conducted into the thermal and hydraulic performance of an air-cooled channel, equipped with a combination of baffles and winglets. Specifically, hollow trapezoidal baffles of a wavy top surface and three sets of winglets have been selected from the literature as the basis of further numerical optimization. First, the spacing between the selected baffles is optimized in both longitudinal (L/PL = 0.2–0.8) and transverse (S/Pt = 0.2–0.8) directions. Artificial neural network modeling (ANN) is employed to predict the optimum design condition and visualize the impact of individual input parameters. The Bayesian regularization algorithm is utilized in conjunction with the backpropagation technique to determine the ideal size of the ANN. Then, delta winglets are introduced between baffle rows with different shapes, arrangements, and orientations. In this study, air serves as the working fluid and is examined across a range of Reynolds numbers (3.8 × 103 ≤ Re ≤ 2.4 × 104). By leveraging numerical simulations and machine learning, the integration of baffles and winglets offers a promising passive cooling technique. For baffles, longitudinal length-to-pitch ratio L/PL = 0.2 and transverse width-to-pitch ratio S/Pt = 0.4 provide the best performance as TEF value reached 1.68 achieving an average gain of 35.9 % compared to the non-optimized one in literature. A comparison of the different sets of winglets demonstrated that delta winglets at a 30° orientation angle provide the best performance. A single row of winglets provided an additional gain of 10.4 % in TEF average value compared to the case without winglets. Two rows of the optimum winglet shape are then optimized in aligned and staggered arrangements at different spacing ratio (PW/PL = 0.3–0.7). TEF attains its peak value of approximately 1.81 at PW/PL = 0.5 in aligned configurations, reflecting an overall 54.7 % improvement compared to the base case.