Reasoning models have changed the AI ball game, says NVIDIA’s Jigar Halani
At DevSparks 2025, Jigar Halani, Director - Solution Architect & Engineering, NVIDIA, noted that artificial intelligence is no longer a futuristic idea—its reshaping industries in real time.
At DevSparks 2025, Jigar Halani, Director, Solution Architect & Engg at NVIDIA, laid out a bold vision for how artificial intelligence, particularly agentic AI, will reshape industries across the country.
During the talk, Halani drew examples from a series of real-life examples to support his statement, where even complex services, such as fingerprint matching for Aadhar verification, are being streamlined by an AI pipeline powered by Convolutional Neural Networks (CNNs).
AI is also tackling India’s language barriers. The country has around 22 scheduled languages and more than 230 dialects that have acted as barriers to digital inclusion. Halani highlighted that AI models can now provide machine translation and speech technologies across languages.
One of the key points highlighted by Halani was the rise of reasoning models, which allow AI to handle constraints and logic in ways earlier systems could not. This ability, he argued, is especially relevant to India’s judiciary, where more than 52 million cases remain pending. Multi-agent AI systems could help judges and lawyers process evidence, summarise documents, and draw from precedents in multiple languages and formats.
Today, NVIDIA has developed tools and frameworks to power this technological shift, which include its NeMo and AIQ stacks for building agentic AI pipelines, optimised versions of open-source models like Llama and Mistral AI as well as high-speed retrieval tools that accelerate reasoning.
From digital twins to autonomous robots, the goal is not just efficiency, but collaboration between humans and machines. “Every single industry that we live with today is going to transform 100% in a new shape completely,” Halani said.
