Lightspeed-backed Yellow.ai to launch LLMs in Indian languages for customer-service functions

“Our strategy will be to continue building more language-specific LLMs on top of foundational models such as Llama-2, based on customer needs. The reason for choosing Llama-2 is because of its robustness in a wide range of downstream tasks,” Khan told Moneycontrol.

The firm recently launched Indonesia's first LLM called Komodo-7B. Trained on 7 billion parameters, this model helps enterprises to reach a wider customer base by offering direct translations between English and 11 regional languages of Indonesia.

“With a less than 1 percent hallucination rate and an average response time of 0.6 seconds, these in-house LLMs are designed to meet the rigorous standards set by enterprises, ensuring secure, precise, and personalized customer interactions,” Khan said.

Founded in 2016 by Raghu Ravinutala, Jaya Kishore Reddy Gollareddy and Khan, Yellow offers AI-first customer service automation to enterprises and has offices across six countries.

Yellow.ai has successfully deployed over 150 generative AI bots for enterprises. Its customer base includes 1100 enterprises, including Hyundai, Volkswagen, ITC Ltd, and Asian Paints.

Also read: India's LLM race is heating up! Here's a look at who's building what

Yellow's Indian language launch comes at a time when several companies are developing Indian language LLMs for enterprise, as well as customer, use cases. Investors are also betting big on this segment.

AI startup Sarvam AI recently launched its first commercial voice-to-voice endpoint tool which will help businesses across voice-related functions like customer support.

Sarvam was the first Indian AI startup to raise $41 million in its Series A funding round led by Lightspeed Ventures with participation from Peak XV Partners and Khosla Ventures in December of 2023.

Similarly, Krutrim AI launched by Ola’s founder Bhavish Aggarwal supports chat functions in six Indian languages.

Yellow’s native language LLMs will help Indian brands that are targeting Tier 2 and Tier 3 cities.

For example, when customers enter search queries in their native language, it accurately interprets and provides relevant product details, improving the shopping experience. Such LLMs can handle customer service inquiries that involve a mix of English and regional languages or even local colloquial words.

“This ensures that the platform can engage effectively with users, understanding and responding to their queries in a natural and relatable manner,” Khan added.

The firm is also working on voice-based LLMs for customer service functions.

“The way voices are being generated and how it is developing human-like conversation is fascinating. We are working on something for customer service functions”  Khan told.

The California-headquartered startup is presently at around $40-50 million ARR as of  FY24, and aims to reach $100-150 million ARR in the next 2-3 years. The firm has raised more than $100 million in funding and counts Lightspeed, WestBridge Capital, Sapphire Ventures, and Salesforce Ventures as its investors.

“In the last three years, we have seen 12-13x growth in our revenue, and 30% of revenue comes from Indian customers…we are close to profitability and the focus is to acquire more customers. We are not looking to raise more funds at the moment,” Khan said.