NER solutions tend to be domain specific, using either encoded word lists in the form of signals (signs) and / or the internal letter structure of the named objects.
Names in the UK and the USA are similar to each other, as well as to other European names, as many of the popular US names come from non-English European immigrants. Indian names vary widely, and commonly used lists of first names / surnames and study materials generally do not contain much or any data specific to India.
To solve your problem, you will need to train a specialized NER model based on data suitable for your specific domain.