Product · Research · Open Source

Open research.

Original products.

Kenpath Labs is where ideas become products. We publish open models like Svara and ship commercial software like KenTraq — built by one small team that owns the whole stack.

Currently shipping —Svara · OSSKenTraq · Product
— What we've built

Two products. One small team.

We work on one open-source thread and one commercial thread at a time — both fully owned, end to end, by the people building them.

OSS·Open Source

Svara

Multilingual speech, native to Indic voice.

An open-source family of speech models built for the languages, accents, and emotional range of India and beyond. Trained in the open, shipped with weights, free for the community.

  • 22+ Indic languages with emotion and prosody
  • Apache-licensed weights, open training notes
  • Built for low-resource hardware and edge inference
500K+
Downloads
22+
Languages
Apache 2.0
License
PRODUCT·Commercial

KenTraq

Know what brands are saying — before it matters.

A live brand intelligence platform that monitors every competitor post, extracts 15+ signals per post with AI, and surfaces market moves in real time. Built for marketing and strategy teams.

  • 15+ AI-extracted intelligence signals per post
  • Updated every 15 minutes — across 6 platforms
  • Auto-generated weekly campaign reports
15 min
Update Cycle
15+
Signals / Post
100+
Languages
500K+
Svara model downloads
22+
Indic languages supported
2
Products in active use
100%
Open weights, open methods
— How we work

A lab wired to a product company.

Kenpath Labs is not a research org pretending to ship, and not a product shop pretending to do research. It's both, on purpose — and the work is structured around that.

Open by default

Research and models we publish ship with weights, methods, and reproducible notes. The default is open; closed is the exception.

Built from where we are

Our research starts from Indic languages, Indian voices, Indian deployment constraints — then generalises outward. Not the other way around.

Research that ships

Labs is wired to a product company. Every research thread is pressure-tested against real teams using it in real workflows.

Small, deliberate teams

We stay small on purpose. Each product is owned by a tight team that can hold the whole stack — model, infra, and interface — in their head.

🤗 Open Source

The open work is the
point, not the preview.

Svara is what we put in the world for everyone. The weights are on Hugging Face, the code is on GitHub, the methods are documented — and the same models power features inside our commercial work.

500K+
Svara downloads on Hugging Face
22+
Indic languages with prosody control
Apache 2.0
Permissive license — weights included
— Timeline

Short history, long bets.

2023

Kenpath Labs takes shape

Started as a small product and research thread alongside the Kenpath team — for long-horizon bets that don't fit a services calendar.

2024

Svara v1 released

First open-source multilingual TTS family — 22+ Indic languages, emotion-aware synthesis, Apache-licensed weights.

2025

KenTraq launched

Brand intelligence platform built on top of in-house multilingual analysis stack. First commercial product from Labs.

Now

Svara v2 + new bets in motion

Working on the next generation of speech models and incubating two unannounced product threads.

— Get in touch

Drop us a note.

Building something at the edge of language, voice, or product intelligence? We're a small team and we read every message.

Direct lines

If you'd rather skip the form, these all reach a human.

Looking for services or partnerships? Head over to kenpath.ai.