What are the key steps to building a GenAI system that rewrites a user's resume in real time under 2 seconds?

1. Understand latency budget across all steps. 2. Pre-process the resume once to save parsing time. 3. Quickly extract job description signals using a lightweight classifier. 4. Create a tight prompt template for the LLM. 5. Stream the response to give perceived speed. 6. Rewrite sections in parallel for efficiency. 7. Cache common rewrites to avoid unnecessary regeneration.