Our Story

Cutting-Edge Voice AI.
Built First for Nepali.

Emotion-aware, full-duplex voice AI - technology usually reserved for English - built first for Nepali, and every low-resource language like it.

Mission

Voice AI Doesn't Stop at English

The frontier of voice AI - emotional nuance, full-duplex conversation - has been built almost exclusively for high-resource languages. We're bringing that same frontier to Nepali.

Not a stripped-down model for "the rest of the world" - the same cutting-edge capability, built first for a language that needed it most.

Why Now

The Tools Finally Exist

Fine-tuning open-source models - instead of training from scratch - makes frontier-grade voice AI achievable without a frontier-grade budget.

That's what makes emotion-aware, full-duplex Nepali voice AI possible today, in a way it wasn't a few years ago.

What Makes It Cutting-Edge

Emotion and Duplex, Not Just Transcription

Most voice AI for low-resource languages stops at speech-to-text. Ours doesn't.

Sentiment-Adaptive Tone

The voice model reads emotional context and adapts its tone in real time - not a flat, robotic read.

See the Platform

Full-Duplex Conversation

Agents listen and speak in the same live session, handling interruptions and backchannel the way people actually talk.

See Full Duplex
Our Defensibility

What Makes OsharaAI a Moat

01

Emotion-Labeled Nepali Data

Emotional-tone and duplex-conversation data doesn't exist off-the-shelf for Nepali. We built the dataset that makes it possible - a model trained on Western internet data has nothing to fine-tune from.

02

Local Trust & Relationships

First-mover trust with the Government of Nepal and local financial institutions - a distribution advantage model quality alone can't buy.

03

Capital Efficiency

Fine-tuning open-source models - instead of training emotion and duplex behavior from scratch - ships frontier capability at a fraction of the cost, and we pass the savings on.

04

Geographic Replication

Win Nepali, and the same emotion-aware, full-duplex playbook replicates for Hindi, Tamil, Sinhala, and every other low-resource language like it.

The Team

Deep Voice AI Expertise, Applied Where It's Needed Most.

Our founders bring hands-on experience shipping production voice AI - speech recognition, emotional tone modeling, and real-time duplex systems - now applied to the language that needed it first.

S

Co-Founder & CEO

Product & Strategy

10+ years building enterprise technology products in global and South Asian markets

Deep experience in AI product strategy, go-to-market, and stakeholder engagement

Led technology initiatives for enterprise clients across Nepal, the US, and South Asia

A

Co-Founder & CTO

AI Research & Engineering

R&D background in speech recognition, emotional-tone modeling, and real-time duplex systems for low-resource languages

Architected production-grade AI systems with high-security, low-latency constraints

Hands-on experience fine-tuning open-source LLMs, voice, and video models for enterprise contexts

AB

Advisory Board

Been There. Done That.

We're backed by advisors who've built and scaled businesses like this - raised funding, run enterprise sales, grown teams in tech and emerging markets.

Serial entrepreneurs with exits and successful funding rounds in enterprise software

Operators who've run data, partnerships, and go-to-market for AI and SaaS companies

Investors themselves, so they know exactly what it takes to get from pilot to Series A

Hear What Cutting-Edge Nepali Voice AI Sounds Like

Book a live walkthrough, or explore full-duplex conversation for yourself.