Kathmandu · South Asia

The substrate for trustworthy AI.

Loktta AI builds the evaluation, agentic, and governance infrastructure that lets enterprises and governments deploy AI they can actually trust — in the languages of the region, on infrastructure they control.

Start a conversation → What we build

evaluated · logged · governed

What we build

One platform, three layers of trust.

Frontier models exist. What's missing is the layer between a raw model and a production system that can be trusted with real work. That layer is what we build.

Evaluation & red-teaming

Independent evaluation, adversarial testing, and multilingual benchmarking for model labs and enterprises — with depth in South Asian languages that frontier platforms treat as second-class.

BenchmarkingRed-teamingRLHF

Agentic infrastructure

Build, deploy, and govern AI agents on infrastructure you control. Regional-language support, human-in-the-loop checkpoints, and observability on every agent run.

AgentsObservabilityOn-prem

GovTech & governance

Locally hosted AI infrastructure aligned with national AI policy — audit logging, data sovereignty, and explainable decisioning for public-sector deployment.

Audit trailsSovereigntyCompliance

Why now

The region is moving from policy to implementation.

Policy has become practice

National AI policies across South Asia have shifted from framework to implementation. Governments need partners who can build, not more whitepapers.

Funded demand is arriving

Development finance institutions are financing digital transformation across the region — with real budgets and a need for on-the-ground implementation.

A gap no one fills

The region's talent is exported as services. No one has built the independent evaluation layer or the sovereign infrastructure that trustworthy AI requires — locally.

Language is underserved

Hundreds of millions speak languages that frontier systems handle poorly. Quality and safety in those languages is measured here, not assumed.

The name

Named after paper that lasts a thousand years.

लोक्ता

Lokta is a Himalayan shrub whose bark has been made into paper in Nepal for over a thousand years. Government records, sacred texts, and land deeds written on lokta paper still survive in temple archives today — legible, intact, trusted across centuries.

That is exactly what we build for AI: the durable substrate on which trustworthy systems are recorded, evaluated, and governed. The head-stroke over our name is the shirorekha — the line that crowns Devanagari script — ending in the torn deckle edge of a handmade sheet.

Who we serve

Built for the people deploying AI in the real world.

Enterprise

Banks & telcos

Multilingual agents and evaluation for regulated, high-stakes deployment.

AI labs

Model providers

Independent red-teaming and regional-language benchmarks for frontier and regional labs.

Public sector

Governments

Sovereign, auditable AI infrastructure for citizen services and modernisation.

Development

DFIs & INGOs

Implementation partners for funded digital transformation programmes.

Founder

Built by an operator who's shipped AI platforms at scale.

Nimesh Kumar

Founder

Nimesh is a product leader who has built enterprise generative-AI platforms at Scale AI, AWS, and Google Cloud — turning bespoke AI projects into reusable platform infrastructure used by hundreds of enterprises. He founded Loktta AI to build that same infrastructure where it's most needed and least served: South Asia.

Connect on LinkedIn →

Let's build trustworthy AI, together.

Whether you're a lab, an enterprise, a government team, or a development partner — we'd value a conversation.

hello@loktta.ai →

or reach the founder directly — nimesh@loktta.ai