AI4Bharat, Josh Talks Launch India AI Evaluation Platform
The platform will test AI across Indian languages, accents and real-world workflows, covering individual models, complete applications and deployed systems.
Topics
News
- Temasek, Seraphim Back Pixxel in $100 Million Round
- UN Rights Chief Warns Advanced AI Could Threaten Humanity
- Tata-owned JLR to Cut About 4,000 Bobs Amid Tariff, China Pressure
- Anthropic Walks Away From $6 Billion Decart Deal
- Hyundai India Targets 20% Women in Executive Workforce by 2030
- Security Breach Drains $320 Million in Bitcoin From Liquid Network
AI4Bharat, a research lab at IIT Madras, and Josh Talks AI have expanded Voice of India into a broader platform for evaluating artificial intelligence systems under Indian operating conditions.
The organizations said on Tuesday, August 4, that the platform would help government agencies, companies, researchers and developers compare systems across Indian languages, accents, dialects and deployment environments.
Voice of India was previously introduced as a benchmark for automatic speech recognition. That project uses 536 hours of unscripted telephone conversations from 36,691 speakers across 15 languages and 139 regional clusters.
The expanded platform is intended to cover voice systems, large language models, document processing, safety and cultural assessments, and Indian legal reasoning, according to the organizations’ press statement.
It will use a three-stage framework. The first stage will test individual models and technical components. The second will assess complete applications under operating conditions, while the third will conduct periodic audits after deployment.
The approach is intended to measure whether systems that perform well in controlled tests remain reliable when used by people with different languages, accents and communication styles.
“Evaluation determines what gets built,” said Mitesh Khapra, a professor at AI4Bharat. He said benchmarks should reflect India’s linguistic diversity as AI systems become more widely used in public and commercial services.
The organizations also announced a finance-focused benchmark covering banking workflows. They did not disclose which models would be tested, when results would be released or how organizations could access the broader evaluation platform.
Supriya Paul, co-founder of Josh Talks AI, said India could contribute to global AI development by building evaluation infrastructure rather than competing only to produce the largest models.
The platform is designed to give developers a common testing framework and provide organizations with independent information before they select or deploy AI systems.
