Our philosophy
Intelligence should serve the pilgrim,
not the platform.
KumbhLabs exists to prove that decentralized, agentic AI can serve tens of millions of people with the same dignity it would offer one.
Where we come from
Born at the intersection of the world's oldest gathering and its newest technology.
No centralized system, however well-funded, was ever designed to serve a population this large, this briefly, this respectfully. The Kumbh is where that assumption breaks in public.
Our founding thesis is simple: citizens are becoming AI-savvy faster than their cities are becoming AI-ready - and that gap is where either empowerment or exploitation gets decided. KumbhDoot is the first large-scale test of the other outcome: a voice-first, multilingual companion built on open protocols rather than a closed platform.
What we believe
Four convictions that shape everything we build.
01
Power should stay with the people an agent serves
An agent that knows someone's health, money, family and standing with institutions holds real power over their life. We build on open, auditable foundations - not closed platforms - so that power sits with the person, never with whoever happens to own the model.
02
Dignity is a design requirement, not a feature
A sixty-year-old Bhojpuri-speaking pilgrim should be able to find a place to stay, a warm meal, a medicine reminder and the safest ghat - entirely by voice, in her own language, without depending on anyone. Systems that require literacy, English or a smartphone-savvy relative do not include people; they move the burden of dependency.
03
Decentralization beats brute force at extreme scale
Tens of millions of people over a few weeks cannot be served by one command centre. Personal agents, an orchestration layer and an open marketplace of providers let thousands of participants coordinate without any single owner of the whole system.
04
A pop-up city is a laboratory for every city
The Kumbh is temporary, but the problems it forces us to solve - safety, crisis response without data silos, mobility without centralized fleets, care without interoperability mandates - are permanent everywhere. Everything we ship for the Mela is designed to outlive it.
How we work
Dimensional AI, not screen AI.
We are explicit about what we are not building: another chatbot, report generator or creativity tool. We build AI that acts in the physical world and reasons about population-scale dynamics.
Research
Agentic AI & edge intelligence
Autonomous agents that serve specific functions and learn from local context, with inference pushed to the edge so the system keeps working even when connectivity does not.
Trust
Computational privacy
Data sovereignty by default - confidential computing and split learning let institutions and providers collaborate without pooling raw data.
Coordination
Self-orchestration
Agents coordinate dynamically across sensors, people and organisations, reallocating resources as density and needs shift in real time.
Fairness
Incentive design
Economic frameworks that reward participants - down to a single flower seller - for contributing to the network, rather than extracting value from them.
Human factors
CrowdUX & population modelling
Agent-based models calibrated to real crowd behaviour, paired with interfaces co-designed with the people who will actually use them.
Our values in practice
What we optimise for when no one is watching.
Open over closed
We build on open protocols, because infrastructure owned by no one is infrastructure that serves everyone.
Systematic safety over situational luck
At this scale, hoping nothing goes wrong is not a strategy. Monitoring, health coordination and help pathways are built in from day one, not bolted on after an incident.
Every language is a first-class language
A Tamil family arriving at midnight and a Bhojpuri-speaking grandmother deserve the same quality of help. Multilinguality is a requirement, not a localisation afterthought.
Build for the festival, design for the city
Anything we validate at the Kumbh is engineered to generalise to any dense population facing the same coordination problems.
“Previously siloed stakeholders - care, mobility, hospitality, crisis response, payments and governance - connected into a single, self-orchestrating civic fabric.”
City-Stack research on decentralized AI