Interview | ‘A Historical Milestone’: How Aakhor AI is bringing Assamese into digital age
Guwahati: In a decisive step toward strengthening regional language computing, brothers Indranil Talukdar and Kabyanil Talukdar have launched Aakhor AI, a software suite built to integrate the Assamese language into the rapidly evolving ecosystem of artificial intelligence.
At the centre of this innovation is “Aakhor AI Voice Typing,” a real-time speech-to-text system that converts spoken Assamese into written script across mobile and desktop platforms. While voice recognition has matured globally for dominant languages, this is the first time such advanced AI-driven linguistic modelling has been designed specifically around the structural and phonetic intricacies of Assamese, including its script conventions and conjunct consonants (juktakshars).
For many observers, this is more than a software upgrade—it represents a technological inflection point for Assamese, ensuring its functional relevance in an AI-dominated digital future.
Krishnakshi Sharma spoke to founder Indranil Talukdar about the origins of the idea, the development process, and the broader implications of Aakhor AI.
What inspired you to develop Aakhor AI, and how did the idea evolve into a full-fledged platform?
The inspiration was rooted in a very practical frustration. I work in IT and spend much of my time programming. At the same time, I write articles for newspapers. Initially, I tried drafting them on my phone, but typing long-form Assamese text on a mobile device was extremely time-consuming. So I shifted to a desktop keyboard, thinking it would be easier.
Instead, I ran into another barrier. Typing Assamese efficiently on a standard desktop setup was slow and cumbersome. The biggest difficulty was handling conjunct consonants—juktakshars—which require multiple key combinations and disrupt writing flow.
That was the turning point. I realised that while global languages were benefiting from technological advancements, Assamese typing tools had remained virtually unchanged for nearly 25 years. There was no innovation, no optimisation, no integration with emerging technologies.
Around 2019, although AI was not yet the global buzzword it is today, I had been independently learning about machine learning and automation. I began exploring whether AI could solve these typing inefficiencies.
Within one or two months, I developed a working prototype of an AI-powered Assamese keyboard. However, we chose not to release it immediately. My brother and I used it ourselves extensively, tested it rigorously, and invited feedback from a small circle of users. We wanted refinement before public deployment.
By 2021, we were confident enough to launch it as Aakhor AI’s first product. Initially, it was built for Windows. Subsequently, we expanded support to macOS and Linux.
As the user base grew, we kept innovating. We integrated a real-time spell checker to assist with sentence formation and linguistic accuracy. We also incorporated an Anglo-Assamese dictionary to improve lexical accessibility.
After briefly pausing to focus on other professional commitments, we returned in 2023 to build a browser-based version, making the tool more accessible without installation barriers.
In 2024, we began development of the voice-operated keyboard. That project required deeper AI modelling, speech dataset generation, and phonetic mapping specific to Assamese pronunciation patterns. We completed it and officially launched the voice typing feature in January 2025.
What is next on Aakhor AI’s roadmap?
Our development priorities are driven directly by user feedback. One of the most consistent requests we receive is for iPhone support. We already support Windows, Linux, macOS, and Android, so launching an iOS version is now our immediate priority.
Another major request is a built-in “Notes” function within the app, allowing users to save drafts directly instead of copying and pasting text elsewhere. This feature already exists on our website, and we are currently working to integrate it into the mobile application.
Beyond application-level features, we are planning something more foundational: a dedicated Assamese Unicode font family. For nearly three decades, users have struggled with inconsistencies and rendering limitations in existing fonts. There is still no Unicode font that fully satisfies aesthetic, typographic, and technical expectations for Assamese.
We believe developing a high-quality font ecosystem will be transformative. It will not only improve readability but also accelerate digital publishing, education, and media workflows. That remains one of our long-term strategic goals.
How large is your team at present?
The core technical team consists of four members. My brother and I lead the programming and architecture, supported by two additional developers.
Beyond that, we have over 20 dedicated volunteers who contribute to data generation, testing, linguistic validation, and corpus development. Their role is critical, especially in training AI systems where high-quality language datasets are essential.
With AI expanding rapidly, there are growing fears about job displacement. How do you see this concern?
At this stage, I do not believe AI poses an immediate existential threat to human employment. In the software industry, productivity has undoubtedly increased. Tasks that previously required five developers can now sometimes be handled by two, thanks to improved frameworks and AI-assisted development tools.
However, even if a team shrinks from twenty to five, those five skilled individuals remain indispensable. AI systems still require human design, supervision, critical thinking, and domain expertise.
Technological transformation has always been gradual. The developments we see today are the result of years of research, trial, and iteration. Future breakthroughs will also take time.
Yes, some companies are reducing manpower, but human intelligence remains central to innovation. Moreover, progress must not be confined to computer science alone. We need advancement in physics, biology, neuroscience, and other disciplines to create balanced growth.
A job crisis may be possible in the long term, but it is not an immediate inevitability.
How has user response been since launch?
The response has been overwhelmingly positive. Assam has a strong literary culture—people here value writing deeply. However, traditional writing methods are time-intensive.
We have seen many journalists transition from pen-and-paper workflows to Aakhor AI. Tasks that once took an hour can now be completed in five minutes using voice typing. The keyboard itself has also gained popularity, with users reporting significantly faster writing speeds.
Of course, not all feedback is uncritical. Approximately 99% of users appreciate the productivity boost, but a small minority argue that such tools may encourage laziness or reduce creativity.
I see technology differently. Its purpose is to remove friction and enhance efficiency. We cannot halt innovation because of philosophical resistance. Choosing voice typing over manual typing is like choosing a car over walking long distances—efficiency enables productivity.
Our objective is not to replace human thought but to remove mechanical barriers to expression. That is what Aakhor AI aims to achieve—and we will continue refining it further.