AI Chat Behavior Analytics App

I want to turn every live-chat session on my site into a stream of insights, not just messages. The web app you build will capture conversations in real time, analyse them, and highlight user behaviour patterns that help us improve support, upsell at the right moment, and spot friction points immediately.

The core of the job is data analysis, so the heart of the solution should be an AI/ML layer (Python, TensorFlow, PyTorch, or a stack you’re confident with) that ingests the live text feed, segments it, and runs pattern-recognition models. Think intent clustering, navigation sequence tracking, and anomaly detection rather than simple keyword counts.

I need this intelligence wrapped in a clean web interface—React or Vue is fine—showing real-time charts plus searchable historical logs. The app must accept chat transcripts through standard webhooks or an SDK, process them on the fly, and push alerts whenever a behaviour rule is triggered (for example, repeated confusion signals or high purchase intent).

Deliverables
• Source code for both the AI engine and the front-end dashboard
• Deployment scripts (Docker, CI/CD) that run on a standard cloud VM
• Brief but complete setup & model-training documentation

Acceptance criteria
• Live-chat messages are ingested and displayed within 2 seconds
• Behaviour patterns (at least three: frustration, high intent, drop-off risk) are detected with >80 % precision on our test data
• Dashboard renders in the latest Chrome, Firefox, and Edge without errors

If you’ve built similar NLP analytics tools or have creative ideas for feature engineering on conversational data, let’s make this happen.