Archive · 26 Dec 2025
Startups Teach AI Indian Dialects
Some Indian startups are teaching AI to understand and speak languages that aren't very common online, like Tulu, Bodo, and Kashmiri.
These languages don't have much digital data, so the startups are collecting voice and text data from people who speak them.
They do this through workshops, WhatsApp voice notes, and community engagement.
The goal is to make AI models that can understand humor, idioms, and cultural context in these languages.
This helps preserve these languages and makes AI more useful for people who speak them.
Big tech companies like OpenAI and Google also support some Indian languages, but these startups are focusing on languages that are often overlooked.
Indian startups are building AI models for low-resource languages like Tulu, Bodo, and Kashmiri.
They collect voice and text data through workshops, WhatsApp voice notes, and community engagement.
The goal is to create AI models that understand cultural context, humor, and idioms in these languages.
Challenges include limited digital data, lack of Unicode standards, and funding constraints.
These efforts help preserve linguistic diversity and make AI more relevant to local contexts.
- Who
- Indian startups like TuluAI, Aakhor AI, and KashmiriGPT
- What
- Building AI models for low-resource Indian languages and dialects
- When
- Ongoing, with recent efforts spurred by the success of global AI models
- Where
- India, focusing on regions where Tulu, Bodo, and Kashmiri are spoken
- Why
- To preserve linguistic diversity, ensure cultural representation, and make AI relevant to local contexts
- Languages Focused
- Tulu, Bodo, Kashmiri
- TuluAI Founder
- Amrith Shenava
- Aakhor AI Co-founder
- Kabyanil Talukdar
- KashmiriGPT Creator
- Saqlain Yousef
- Tulu Speakers
- 2 million
- Kashmiri Speakers
- 7 million
- Data Collection Methods
- Workshops, WhatsApp voice notes, community engagement
- Challenges
- Limited digital data, lack of Unicode standards, funding constraints