VP, Head of AI at Krutrim
As the founding Head of AI at Krutrim, Chandra Khatri is building the full AI stack for India, from chips to AI cloud to multilingual foundation models for consumer and enterprise applications. Driven by India's linguistic diversity with over a billion voices, his mission is to bridge the gap between urban and grassroots communities through the development of multimodal and multilingual foundation models. His team recently built "Krutrim", the world's first India-centric multilingual LLM, which outperforms several state-of-the-art foundation models in the Indian context. Before Krutrim, he founded Got-It AI, a pioneer in conversational AI, developing cutting-edge automation products including the accurate Enterprise Language Model Architecture (ELMAR) and the first Hallucination Detection Platform (TruthChecker). He also spearheaded the creation of the world's first fully autonomous Conversational AI, significantly advancing virtual agent deployment. Before his tenure at Got-It AI, he established or led several AI teams at Amazon, Uber, and eBay, and created the Alexa Prize, the first voice-centric consumer-facing large-scale open-domain conversational system, akin to ChatGPT for Amazon Alexa users, built 5 years before ChatGPT emerged. In addition to developing products, he invests in or serves on the boards of cutting-edge technology companies such as ThirdAI and Optivolt.
All Sessions by Chandra Khatri
Everything About Large Language Models: Pre-training, Fine-tuning, RLHF & State of the ArtGenerative AI | All Levels
Generative Large Language Models like GPT4 have revolutionized the entire tech ecosystem. But what makes them so powerful? What are the secret components which make them generalize to a variety of tasks? In this talk, I will present how these foundation models are trained. What are the steps and core-components behind these LLMs? I will also cover how smaller, domain-specific models can outperform general purpose foundation models like ChatGPT on target use cases