Who we are

At Navana, we're building the leading voice AI infrastructure company for India. We own the full stack: speech recognition (STT), speech synthesis (TTS), a small language model, and the orchestration layer that holds them together. Our models are built in India from the ground up and run live inside banks, insurers and lenders, on-prem and in the cloud.

Since 2018 we've trained on real conversations across 12+ languages and 45+ dialects, because the person on the other end of the line is calling from a small town, switching languages mid-sentence, about their money. When something breaks, there's no vendor to point at. It's ours.

The problem you're here to solve

Navana is scaling two fast-growing products: an enterprise AI contact center for highly regulated industries (banks, insurers) and a developer API platform that requires rapid feedback loops.

Core Challenge:

Product growth is outpacing the current ability to test reliability, necessitating a dedicated QA function to identify limits before issues reach production.

This isn't a manual QA role.

You will build the QA function from scratch, define quality standards for both products, write the foundational testing processes, and build the QA team. The role requires a hybrid approach, combining targeted manual testing where judgment is needed with comprehensive release automation production observability.

What will be true because you're here

In your first few months:

• Nothing reaches a customer untested. Both products have a release bar, and it's different for an enterprise rollout and a developer-facing update.
• Load and scale testing runs before every major release, so we ship knowing exactly how much concurrency and latency headroom we have.
• The flows that matter most have automated regression coverage that runs on every release, and manual testing goes where judgment beats scripts.
• On the API platform, monitoring tells us something broke before a developer does, and external feedback becomes triaged, fixed issues on a loop the team can see.

In your first year:

• We have a test strategy for what clients actually depend on: platform performance and stability under real load, compliance on every release, and the outcomes they bought us for. It covers latency and concurrency at scale, uptime across cloud and on-device deployments, and the regulatory checks our BFSI clients answer to. Agent behaviour is part of it, but a smaller part as long as it aids you to deep dive into platform / feature testing.