Misraj AI launches Arabic-first AI ecosystem at AWS re:Invent
Baseer document intelligence tool targets enterprise automation
#SaudiArabia #USA #Arabicfirst - Riyadh-based Misraj AI has unveiled a unified Arabic-first AI stack at AWS re:Invent 2025, Amazon Web Services’ annual technology conference. The new AI stack enables advanced language modeling, dialect processing, document intelligence, and developer access to structured Arabic knowledge systems. The Kawn platform includes large language models, vision-language models, embedding models and retrieval-augmented generation pipelines engineered for native Arabic language understanding across formal Arabic and regional dialects. According to the company, the platform’s Lahjawi dialect engine is capable of recognising more than a dozen dialects with high accuracy.
SO WHAT? - Misraj AI’s new Kawn platform is designed to address a fundamental challenge facing Middle East organisations: the majority of institutional knowledge remains locked in unstructured Arabic documents that existing AI systems, predominantly built for English, cannot process effectively. The launch of the company’s platform is one of a growing number of Arabic-first platforms being brought to market to help enterprises to maximise the value from their’ Arabic data and develop new Arabic-first solutions.
Here are some key points about Misraj AI’s new products:
Misraj AI unveiled the Kawn ecosystem together with its SeamlessAPI at AWS re:Invent 2025 this week, presenting both platforms as foundational components of modern Arabic AI infrastructure designed for enterprise, government, education and healthcare applications.
According to the developer, the Kawn ecosystem provides a comprehensive Arabic-first AI stack including advanced large language models, vision-language models, embedding models and RAG (retrieval-augmented generation) pipelines built to model real linguistic usage across formal Arabic and regional dialects.
Kawn features Lahjawi, a dialect engine capable of recognizing and responding to more than a dozen Arabic dialects with high accuracy and native fluency, engineered for native Arabic language understanding rather than adaptation from English-centric models.
SeamlessAPI extends Kawn capabilities by providing developers with a unified interface to access Arabic knowledge, translation tools, dialect detection and document intelligence, enabling single API calls to analyze Arabic text, summarize content, translate between languages or extract structured data.
Misraj AI simultaneously launched Baseer, a state-of-the-art Arabic document intelligence platform that delivers advanced OCR-to-Markdown conversion, transforming scanned documents and PDFs into clean, structured, AI-ready data whilst preserving layout elements and supporting multilingual content.
Baseer enables organisations to convert legacy archives into structured data, automate compliance and legal workflows, and build Arabic-native knowledge systems, directly supporting digital transformation goals including Saudi Arabia’s Vision 2030 initiatives.
ZOOM OUT - Global AI models have historically underperformed in Arabic language processing due to dialectal complexity and most enterprise platforms are English-first. A growing number of startups are addressing this gap, with Saudi Arabia emerging as a hub for Arabic-first platforms. Arabic language processing represents a persistent challenge for many industries in the Middle East, with most natural language processing systems being optimised for English and other Latin-script languages.
[Written and edited with the assistance of AI]