looking for expert who can fix my AI CHAT BOT

My AI chatbot lives on our website and normally handles customer queries well, but lately it fires back irrelevant or empty replies at random. The pattern is unpredictable: a question it answers perfectly one moment will trigger a blank or nonsensical response the next. The underlying model itself seems solid, so I suspect the fault sits in either the dialogue-management logic, session handling, or an overlooked edge-case in the NLP pipeline.

Here is what I need from you:

• Diagnose the exact trigger for the “random bad reply” behaviour, whether it stems from state management, intent classification, caching, or an integration glitch.
• Patch or refactor the affected code so the bot responds correctly every time.
• Stress-test the fix with a variety of inputs and simulated traffic spikes, then hand me a concise report showing before-and-after accuracy.

The bot is already deployed on a production website, so any debugging must be done through a staging branch or with minimal downtime. If you work comfortably with typical chatbot stacks—Python, Node, Dialogflow, Rasa, or cloud NLP services—you should feel right at home. All source files and access tokens will be provided once we agree on a path forward, and I’m happy to walk you through the current architecture on a quick call.

Deliverables are accepted when the bot maintains consistent, contextually correct answers over a 48-hour monitored window and test logs confirm the fix.