Shipped multi-tenant AI calling SaaS — 4,800 lines, full stack
Built end-to-end: ElevenLabs voice + Tata Tele SIP, automated tenant provisioning, billing, observability. Pricing benchmarked against Ozonetel/Exotel/MyOperator.
12ms
Mean response latency
4,880
Lines of production code
60%
Cost vs. nearest competitor
What they came in with.
We needed a real AI calling product for the Indian market — one that handled TRAI compliance, sub-15ms latency on Tata Tele SIP, and multi-tenant provisioning out of the box. Existing US-built tools couldn't deploy here without breaking compliance or quality.
How we worked it.
- 01
Audited every viable Indian SIP trunk provider — landed on Tata Tele after Plivo hit DLT registration friction.
- 02
Built a multi-tenant SaaS architecture: per-workspace number provisioning, isolated context, automated billing.
- 03
Wired ElevenLabs voice synthesis with custom turn-taking logic to keep latency under human-perception thresholds.
- 04
Benchmarked pricing against Ozonetel, Exotel, MyOperator — landed at ~60% of nearest competitor for equivalent features.
What it earned.
Shipped 4,880 lines of production code across frontend, backend, and infrastructure. Mean response latency 12ms. First three customers onboarded via self-serve provisioning. Currently in beta with selective rollout.
Then we should work together.
A 30-minute strategy call. We diagnose where AI actually fits in your business — and where it doesn't. No deck, no pitch, no cost.