top of page
Abstract Graphic Shapes

Revolutionizing Panjabi Community with GenAI

MISSION

Revolutionizing the advancement in Punjabi language modelling by building end to end AI Stack 

We aim to build a unified platform to provide end to end AI stack in Punjabi language in order to provide enterprise solutions, taping consumer end cases and building global network effects

Why we are building this?

BRANDING

Position as a global brand for panjabi community where consumers and businesses interact with each other

PLATFORM

Serves as a platform to build end to end AI stack for vertical integrations across enterprises

WORLDWIDE COMMUNITY

Significant global presence (USA, Canada, Australia, England, UK etc.). !0th Most speaking langauge in the world

PRESERVING CULTURE

Preserve Panjab's rich culture, heritage, history and its literature for future generations by building SOTA reasoning

OUR RESEARCH PARTNERS

AWS activate.png
gcp Partner.png
Salad logo.png
IBM.jpeg
Microsoft startups.png

AI PRODUCT STACK

KAPANI - AUTOMATIC SPEECH RECOGNITION MODEL

Introducing our pioneering Automatic Speech Recognition (ASR) model - Kapani for Punjabi speech recognition, crafted to overcome the scarcity of high-quality voice data. Our specialized ASR technology precisely captures and interprets Punjabi speech, enabling fluent communication for Punjabi speakers globally. With advanced algorithms and deep linguistic insights, we're advancing voice technology to ensure that Punjabi is prominently represented and seamlessly integrated into the digital world.

KAPANI FEATURES

1

LOWEST WER

Upon evaluations with fluers, commonvoice evaluation sets, WER of our model is lower than Google USM which is currently SOTA for Punjabi Audios

2

LOWEST COST PER MILLION CHARACTERS

Model's per million character output cost is much lesser than current SOTA model by Google USM

3

LIVE TRANSCRIPTION

Live audio transcription is available which makes it really useful for practical real world applications

RAAG - TEXT TO SPEECH MODEL 

Our in-house AI-developed Text-to-Speech model sets a new benchmark for Punjabi voice synthesis. Unmatched in expressiveness and emotional depth, it brings Punjabi text to life with unprecedented authenticity. As part of our end-to-end Punjabi AI stack, this TTS technology represents a leap forward in how AI interacts with the Punjabi language.

SAMPLES OF OUR TTS ENGINE

AKHAR - LARGE LANGUAGE MODEL IN PUNJABI

Building foundational LLM - Akhar in Panjabi language to enable solid reasoning capability for panjabi dialects. Upon release, it will be the best known LLM for panjabi community which serves as a platform to solve industry specific uses much more efficiency and accurately.

bottom of page