Solving LLM Limitations & Building a Serverless AI Agent Locally