This work presents a machine learning (ML) enhanced design methodology for Operational Transconductance Amplifiers (OTA) for low-frequency applications. The method integrates the g m/I d with Random Forest Regression (RFR) to enable prediction of optimal transistor geometries. Unlike the g m/I d look-up-table based methods, which require iterative manual tuning, the ML integrated process uses the data from extensive SPICE simulations to train a predictive model. The model achieves an R 2 score of 0.955 and a Mean Absolute Percentage Error (MAPE) of 5.5%. The RFR-optimized OTA is subsequently used in a Capacitive-Coupled Instrumentation Amplifier (CCIA) architecture which utilizes the switched-capacitor configuration to replace large feedback resistors. The ML-assisted approach in the analog design addresses key challenges in low-frequency signal acquisition, including power efficiency, area minimization, and signal fidelity. The simulation results indicate that the OTA achieves an open-loop gain of 65 dB and 90 µW power consumption, while the switch-capacitor-based CCIA delivers a closed loop gain of 53 dB while consuming 150 µW of power. The proposed CCIA achieves a third-harmonic distortion of 14.33 dB and an input-referred noise level of 9.7 mV/√Hz. The overall figure of merit is calculated at 104.9 × 10 9, validating the effectiveness of the ML-assisted methodology.