Welding titanium remains a significant challenge because of its high reactivity with atmospheric gases at elevated temperatures, leading to potential defects. This study investigates the effect of welding parameters in Tungsten Inert Gas (TIG) welding on the quality of titanium welds. Experimental trials were conducted on Grade 2 titanium sheets, varying key process factors such as current, shielding gas flow, and nozzle geometry. Additionally, artificial intelligence-based defect detection model using the YOLO convolutional neural network was applied to evaluate the weld quality. The results demonstrate that optimizing these parameters significantly reduces oxidation and improves weld penetration. The highest-quality weld was obtained using a welding current of 83 A, a shielding gas flow rate of 15 L/min at the weld face, 14 L/min from an auxiliary device, and 3 L/min at the weld root. A 14 mm nozzle with a gas lens effectively minimized surface oxidation, leading to a defect-free weld as confirmed by AI-based detection. The YOLO-based defect detection model achieved high precision and recall for most defect classes, including non-conformance (95.1% / 80.0%), geometric (87.8% / 71.4%), and post-processing defects (100% / 72.1%), with lower performance observed for adjacent (74.8% / 60.5%) and integrity defects (86.4% / 27.3%). This study confirms the potential of integrating AI into welding process evaluation, highlighting the role of shielding gas distribution in achieving high-quality titanium welds.