Multi-Agent Chat Automation

I’m ready to introduce an AI-driven customer-support layer inside my SaaS platform https://freelancerai.vercel.app/ and need a specialist who can architect, train, and deploy a set of cooperating agents that handle live-chat conversations end-to-end. The goal is to move beyond simple keyword replies and deliver responses that show genuine contextual understanding while still keeping the system lightweight and maintainable.

Here’s the current environment: the application is built on a typical React front-end with a Node/Express API and a PostgreSQL database. You’ll plug your solution into our existing WebSocket-based chat widget, so familiarity with real-time messaging workflows is important. I’m open to frameworks such as LangChain, Rasa, or a custom orchestration layer on top of OpenAI/Anthropic models—as long as the final solution lets multiple specialized agents collaborate (e.g., intent detection, knowledge retrieval, escalation) before a single answer is sent back to the user.

Key deliverables
• Agent architecture diagram and tech stack decision rationale
• Working multi-agent backend (containerised) integrated with our live-chat endpoint
• Training pipelines and prompts achieving moderate contextual understanding (70 %+ accuracy on our sample set)
• Admin console or API routes to review, retrain, and update each agent independently
• Deployment guide and hand-off session

Acceptance will be based on: seamless connection to our chat widget, response latency under two seconds for 95 % of requests, and successful handling of at least 80 % of typical support scenarios in our test suite without human takeover.

If you have solid, recent experience launching conversational AI in production and can iterate quickly alongside our dev team, I’d love to see how you’d tackle this.