Multi-Tenant AI Voice Agent Platform

I need a production-ready SaaS platform that lets different client accounts spin up their own AI voice agents in Malayalam, English and Hindi. The agents will answer inbound calls and place outbound ones for customer support first, but they must also be able to capture leads, run outreach campaigns and close basic sales when flows allow.

Voice connectivity must be plug-and-play with TeleCMI, Twilio, Exotel or any comparable cloud telephony number. I expect full IVR support, dynamic call routing, call recording and the ability to fall back to a live agent when confidence drops.

Key components and expectations
• Multi-tenant architecture with role-based access so each tenant only sees its own data.
• NLU, ASR and TTS tuned for the three target languages; switching between them mid-call when required.
• Visual conversation builder, intent/utterance manager and analytics dashboard (calls, transcripts, conversion metrics).
• REST/GraphQL APIs and webhooks that let external CRMs or ticketing tools trigger or receive calls.
• Campaign scheduler for outbound lead outreach, including DNC compliance and retry logic.
• Deployment on cloud (Kubernetes or serverless) with CI/CD hooks so features can roll out without downtime.
• Source code, documentation and a short hand-off session are part of the final package.

Acceptance criteria
1. A demo tenant can be created in under two minutes.
2. Inbound call from a TeleCMI DID is answered in the correct language and routed through at least one branching IVR option.
3. Outbound dialer places 50 concurrent calls without degradation.
4. Dashboards show call logs, transcripts and sentiment scores in near real-time.
5. All three languages achieve ≥90 % intent recognition accuracy on the supplied test set.

Timeline is ASAP, so please outline how you will reach a working MVP quickly and expand to full production features right after. When you respond, share past work that proves you have already built or integrated voice bots, CPaaS services or multi-language NLU; that portfolio will weigh heavily in my decision.