KumbhDoot: A Scale-Ready, LLM-Bounded Architecture for Mass-Gathering Public-Service Assistants

A research paper from arXiv discusses KumbhDoot, an agentic pilgrim assistant designed for mass gatherings like the Kumbh Mela. This system addresses the limitations of conventional Large Language Model (LLM)-based conversational assistants in high-demand, safety-critical, and potentially connectivity-challenged environments. KumbhDoot prioritizes semantic similarity over direct LLM queries, aiming to provide a scale-ready and robust solution for information dissemination in such scenarios.

Why it matters

This research is strategically important because it introduces an alternative architectural principle for AI-driven public service assistants, moving beyond the default reliance on LLMs. This approach could lead to more cost-effective, reliable, and safer solutions for critical information delivery in high-stakes environments, influencing future system designs for public-facing AI applications.

Key insights

• Traditional LLM-centric conversational assistants are costly, slow for emergencies, prone to hallucination, and dependent on connectivity, making them unsuitable for mass gatherings.
• The KumbhDoot system is an agentic pilgrim assistant specifically developed for the Nashik Simhastha Kumbh Mela.
• KumbhDoot's foundational design prioritizes semantic similarity for query processing rather than immediately routing to an LLM.
• The system aims to provide scale-ready, LLM-bounded architecture suitable for high-demand, multilingual, and safety-critical information needs.
• The design addresses challenges posed by mass gatherings, including intense, repetitive, and safety-critical information demands.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2608.07520