Voice agent · For any team that takes calls
Picks up the phone and talks like the people who call.
For support lines, clinics, banks and telecoms. Callers speak English, Tamil or both mixed, and it answers in their language, then logs the call in English for your team.
Caller
இண்டைக்கு எங்கட வீட்டுக் குப்பை எடுக்கேல்ல.
Nobody picked up our garbage today.
The problem
People speak dozens of languages. Most phone lines speak one.
Callers who don't speak English wait for an interpreter, fight menus that don't understand them, or hang up. Interpreter minutes add up, and the long tail of languages is where interpreters cost the most and AI is weakest.
- languages on our roadmap
- 16
Dialects
The same language sounds different from one city to the next. Models usually learn only the standard form.
Slang and code-switching
Real conversations mix languages and scripts mid-sentence. Training data rarely does.
Cultural context
Kinship terms, honorifics, food, faith and humour carry meaning a literal translation misses.
Demo
One example: a 311 call in Jaffna Tamil.
Caller
Uyir agent
Call log
Where it works
Any line that hears more than English.
Government and 311
For City Hall there's Townline: permits, 311, public records, licensing and inspections.
See Townline →Healthcare
Booking and reminders, intake questions, directions and what to bring.
Banks and insurance
Account questions, claim status, card and payment help.
Telecom and utilities
Billing, outages, plan changes and moving service.
Customer support
Orders, returns, deliveries and bookings, for any business.
Community and settlement
Intake, referrals and appointments for newcomers and families.
For your callers
Anyone can call, in the language they speak at home.
Uyir agents answer people the way a neighbour would: in their own language and dialect, on the phone or in chat.
No waiting for an interpreter
Callers speak their language from the first word. No transfer, no hold while a third party joins.
Understood the way they talk
Dialects, slang and switching between languages mid-sentence, like Jaffna Tamil or Tanglish.
A clear next step
A reference number, an appointment, the documents to bring. Confirmed in their language, logged in English for your team.
Own your intelligence
Built on open models a customer can run itself.
01
Our own research
Nila, our Tamil model, is trained from scratch on Tamil, English and Tanglish. The agents use open-weight models that can move onto a customer's own servers.
02
Your data agreement
Today the demos use hosted speech services. The language models are open-weight, so a pilot can run them in a cloud the customer chooses.
03
Gets better where it's used
Each customer's languages and dialects improve with use, under their control.
Pilots
See it working in the languages your people speak.
Businesses, governments and organizations serving multilingual communities: tell us who you serve and the languages they speak. We'll show you the agent working in them and scope a pilot around your calls.