The present research investigates the effect of process parameters on the microstructure, micro segregation, and material properties of the Autogenous Double Pulse Tungsten Inert Gas Welding (ADP-TIG) of Inconel 625. The optimization of the process parameter was evaluated by Response surface Methodology (RSM) Box Behnken method. In this article, the effects of the weld-bead geometry, weld centre (WC), weld interface (WI), and heat affected zone (HAZ) were evaluated through macro morphology, microstructure, SEM/EDAX, XRD, and micro hardness. The outcomes indicate that by employing a combination of high and low frequency pulsing in ADP-TIG, it is possible to regulate not only the dimensions of the weld-bead and penetration depth but also the HAZ. The ADP-TIG process has a low heat input, resulting in finer grains in the microstructure, reduced secondary phases, and a narrower HAZ. SEM images of alloy 625 confirms the depiction of fine equiaxed dendrites. The ADP-TIG process reduces secondary phase formation due to a reduction in Mo and Nb alloying elements. EDAX results also show that the micro segregation in the weld centre (WC) and weld interface (WI) is very less in the dendritic core region because of the lower heat Input. The XRD result confirms the presence of NbzNi, Cr₂Nb, Ni8Nb, Fe2Mo3 and FeNi phases. The maximum microhardness values in WC and WI are 290 HV and 280 HV, respectively. The impacts of ADP-TIG welding technique specifications on factors such as weld bead geometry, WC, WI, and HAZ have been thoroughly examined across a range of process parameters. Further the process parameters of ADP-TIG were optimized using a desirability function to achieve defect-free welds, followed by experimental validation. The desirability function results were then compared with a machine learning (ML) approach. The investigation aimed to assess ML algorithms' predictive capability for weldment outcomes. From the ML modelling results, Support Vector Machine (SVM) using the Radial Basis Function (RBF) kernel was identified as the superior strategy for accurately predicting Heat input, Mo segregation, and Nb segregation. Additionally, Gaussian process regression (GPR) with an RBF kernel was recommended for predicting the Width to Depth Ratio.