Build Journal · AyurDoc

Live transcription in Hindi, Kannada and English: what multilingual consults demand

By Ganesh R Datta. I build healthcare AI end to end: clinical decision support, clinic intake, and the marketing and operations around them.

A patient describes their symptoms in Kannada, the practitioner asks a follow-up in English, the family member adds context in Hindi. That is a normal consultation in Bengaluru.

Why English-only fails here

If the transcription only handles English, the most natural parts of the history, often told in the patient's first language, are the parts that get lost. The draft note then looks complete and is not.

What I prioritised

AyurDoc supports live transcription in Hindi and Kannada as well as English. The aim is a draft note that reflects the whole conversation, written in the structure the practitioner will review.

Practical considerations

• Mixed-language sentences are common, so the system has to handle switching mid-sentence, not just per conversation.
• Clinical terms often stay in Sanskrit or English even when the conversation is in another language.
• The practitioner reviews everything, so the draft should make uncertain passages easy to spot rather than guessing confidently.

For founders building for India

Language is not a localisation task you do after launch. It decides whether the product works in the room at all.

More from the Build Journal

• What building clinical decision support across streams of medicine taught me
• Why Ayurvedic practitioners need their own clinical software
• Designing a clinical note the practitioner signs instead of rewrites
• Building AI that suggests the next question without leading the consult

The full Build Journal · The AyurDoc work · Book a 15-minute call