Global Agriculture

An AI startup called Farm Vaidya (FV) has developed a generative AI voice agent that provides agricultural advice to Telugu-speaking farmers in southeast India through mobile phones. The project, advised by IFPRI, encountered significant language-related challenges during development.
Existing speech-to-text models showed error rates exceeding 50% due to dialects, slang, and noisy farm environments. To address this, FV collected over 30,000 hours of real conversational farm audio across India's 10 major languages, targeting a word error rate below 3%. The team also documented more than 5,000 previously unrecorded vernacular terms for crops and pests by recording village conversations.
The system serves Telugu speakers among India's 90 million farmers and has been tested in Tamil as well. During beta trials, the platform supported over 5,000 minutes of live farmer interactions. The project highlights the importance of developing AI tools in local languages and dialects to bridge information gaps for non-English-speaking and illiterate farmers across India's 13 major languages spoken by roughly 1 billion people.
Bangladesh agri-tech platforms use Bangla apps for farmers
Marathi Made Mandatory for Auto-Rickshaw Drivers in Maharashtra
Chrome adds support for 8 Indic languages in AI features
Telangana digitizes 1.8 lakh manuscripts in 15 languages
93% of Karnataka students choose Hindi as third language
AI Farm Advisor Built for Telugu Speakers in India
