Built by people you will actually talk to
PHOTO
[FOUNDER NAME] — Founder
[Two or three concrete sentences. What you built, where, and why it makes you credible on multilingual voice AI specifically. Name systems and outcomes rather than years of experience. If your background is not in insurance or voice, say what it is and let the benchmark carry the domain credibility — a thin bio next to real numbers is fine; an inflated bio next to no numbers is the worst combination.]
Why we are doing this
We build small language models for regulated voice conversations. We started because the same thing kept happening: a company with obvious call volume, an obvious use case, and a compliance function that could not approve any of the available options.
The thesis is narrow. Most of what happens on a regulated call is bounded work that does not need a frontier model — and on code-mixed Indian speech, a model fine-tuned on real calls should be better than a frontier model rather than merely cheaper. We intend to prove that with published numbers rather than argument.
How we work
Small team, direct contact. You will talk to the people building the system, not an account manager. We tell you when the answer is that you should keep using a hosted API.
Structure
Vaakya is a Engineering and data work happen in India. [Confirm final wording with your counsel before publishing.]