Delhivery announced that it is advancing the development of digital mapping solutions tailored specifically to Bharat's complex geography and unique addressing systems with NVIDIA AI.

Announced at the India AI Summit, Delhivery aims to leverage NVIDIA accelerated computing, NVIDIA CV-CUDA, and NVIDIA Nemotron open models to design highly accurate, scalable and locally relevant maps to support India's growing digital economy by unlocking the potential of Delhivery's petabytes of proprietary data accumulated over several years. Key capabilities being evaluated include address disambiguation to distinguish between phonetically similar addresses and localities, contextual inference to decipher unstructured descriptions, and predictive sequencing to optimize last-mile routes in dense, unplanned urban areas.

India's geographic diversity, rapid urbanization and unique addressing systems present a distinct challenge for conventional mapping tools built for other nations. Addresses in the Indian subcontinent are often unstructured, relying heavily on community landmarks, variable sequencing, and local vernacular rather than standardized global coordinates. Through this collaboration, Delhivery aims to harness NVIDIA's open models to tackle critical hurdles unique to India.

“Bharat needs mapping solutions that operate at scale and can understand the pulse of its streets. Standard global mapping tools often struggle with nuances of Indian addresses. By combining Delhivery's dataset of billions of shipments with NVIDIA's accelerated computing, we are exploring a location intelligence layer to power India's digital growth,” said Kapil Bharati, Co-founder & Chief Technology Officer, Delhivery.