Fix slow startup: defer heavy ML library imports (sentence_transformers, litellm)

Improve TalentAI provider endpoint validation

Add support for TalentAI provider

deploy