The Making of Sarvam-CLI — How I built India’s 2nd Sovereign AI Coding Assistant

It started with a Small Copilot Window in early 2023…

I started learning AI in early 2023. Not in a classroom, not through a course with a certificate at the end of it — but by simply opening up Copilot and beginning to type.

For most people, AI was just a curiosity, a productivity nudge, a “nice to have.” For me, the benefits were immediate and personal. I have had cerebral palsy since birth. The friction between what my mind wanted to type and what my hands could physically execute had always been real. Every line of content I produced carried a small physical tax (on myself). And here was a tool that could take an intention and turn it into a ready to send draft — that could meet me halfway, and sometimes more than halfway.

The first time Copilot completed a paragraph I had only prompted for, something clicked. This wasn’t a gimmick. This was leverage. This was a machine that narrowed the distance between thinking and doing, and for someone like me, that distance has always been the whole game.

So I didn’t put it down. I kept going…

What became obvious very quickly was that AI doesn’t sit still. The tool I used in the morning wasn’t quite the same tool I’d be using a month later. Models improved. Capabilities expanded. Entire categories of products appeared that hadn’t existed weeks before. AI was — and is — ever-evolving. If I wanted to actually understand it rather than just use it, I couldn’t treat it as a static skill that you’d learn one day — only to forget it the next. I had to treat it as a moving target. So I started keeping track. Newsletters, release notes, demos, tutorials, videos, reddit threads, papers I half-understood and reread until I understood them better. I became a student of the pace rather than individual pieces of technology.

The Quiet Frustration That Pushed Me Forward There’s a particular feeling that comes from working inside an enterprise environment. The tools are fixed. The stack is decided. The approved list is the approved list, and innovation happens on someone else’s timeline. There’s safety in that, and there’s also a slow, creeping boredom. Progress felt incremental when the whole world outside felt like it was sprinting. I’ll be honest about the real engine underneath my curiosity, because I think a lot of people quietly share it and rarely say it out loud: fear. The fear of being left behind. The fear of waking up one day and discovering that the ground had shifted and I hadn’t moved with it. The fear of losing my job — not to a person, but to a wave I had decided to watch instead of ride.

Fear is an underrated teacher. People love to talk about passion and curiosity as motivators, and those matter. But fear got me out of the chair on the days the other two didn’t show up. Fear made me open the laptop on a tired evening. Fear made “I’ll learn it eventually” turn into “I’m learning it right now.” I’m not ashamed of admitting that. I used it to my advantage.

But fear alone makes you defensive. It makes you learn just enough to survive. What I needed was something that would turn that defensive crouch into a forward lean. And that came from an unexpected place.

A Few YouTube Videos That Rearranged My Thinking The real push — the shift from anxious learner to obsessed builder — came from a handful of YouTube videos. I don’t even remember all of them precisely now, but I remember the arguments they made, because those arguments resonated. They reorganized how I saw my own situation. It went something like this:

AI offers a level playing field. For the first time, the gap between a large team with deep resources and a single motivated individual was shrinking. The assistant that helped a senior engineer at a big company was, more or less, the same assistant available to me. The tools people have access to are being democratized. Capability is being handed out at the crossroads.

GitHub is a playground. Not a résumé booster. Not a portfolio to be carefully curated and feared. A playground — a place to build in public, to try things, to break things, to ship something rough and learn from it and ship again. The mental reframe mattered enormously. A playground is somewhere you go to play. You don’t need permission to play.

Those who start building will thrive. Everyone else will struggle to survive. This was the line that lodged itself deep in me and wouldn’t leave. It wasn’t about talent. It wasn’t about credentials. It was about a single binary choice: do you build, or do you just watch? The dividing line of the next decade, the videos argued, wouldn’t run between the gifted and the ordinary. It would run between the builders and the spectators. I had spent the better part of a year being an excellent spectator. I knew the landscape. I tracked the releases. I could tell you what was new. But I wasn’t making anything with it.

That stopped.

A Love Affair With AI Tools and GitHub What followed, I can only describe as a love affair.

I fell for the tools and I fell for the platform (i.e. Github). Every new AI capability became something to test by building. Every idea that wandered through my head became a repo. The loop became addictive in the best way: imagine something, prototype it with AI assistance, push it to GitHub, learn what broke, repeat. The friction that had always defined my relationship with ideas turning into code — the physical tax, the slow grind — was now buffered by tools that absorbed the tedious parts and left me with the interesting parts.

I built constantly. I built things that worked and things that didn’t. I built things nobody would ever use and a few things I was quietly proud of. I wasn’t building for an audience. I was building because building had become the way I thought.

By the end of June 2026, I had done close to 70 projects on GitHub.

Seventy. In one year. A man of 51 with cerebral palsy. One working in a non-technical role. But more importantly, one who started out by simply not putting Copilot down.

I don’t share that number to impress anyone. I share it because of what it represents: the compounding return of choosing to be a builder instead of a watcher. Those 70 projects didn’t happen by Accident or by flook. Those 70 projects couldn’t have happened overnight — even with AI. None of those 70 projects required concent or permission from anyone. None of them required me to be the smartest person in any room. They only required one thing — showing up consistently, and using the leverage that was now available to everyone — including me.

One Blog and a Question About Sovereignty About two weeks ago, I wrote a blog on Digital Sovereignty in Europe.

It’s a subject that pulls at something deeper than technology. Digital Sovereignty is the idea that a nation should have meaningful control over its own digital infrastructure — its data, its models, its tools — rather than depending entirely on systems built and governed elsewhere. Europe has been wrestling with this question seriously, trying to chart a path where it isn’t simply a customer of technology made on other continents but an owner of technology made on its own.

Writing that piece did something to me. It planted a question that wouldn’t go quiet. If Europe was thinking this hard about digital sovereignty — what about India?

I wanted to do something about Digital Sovereignty in India. Not just write about it. Do something. The builder reflex, by now, was automatic. A question wasn’t satisfying anymore unless it turned into a repo.

So I started exploring. And the natural place to look was at the people already doing the hard, foundational work: Sarvam.ai.

The Integrations Page and the Eureka Moment I started digging into what the people at Sarvam were building. Sovereign AI for India — models built with India in mind, the kind of foundational work that digital sovereignty actually requires. The more I read, the more I respected it. Here was a serious answer to a serious question.

And then I landed on their integrations page. I want to describe this moment carefully, because it’s the hinge of the entire story.

I was looking at the tools that connected to this sovereign AI. And the tools were… ChatGPT. Cursor. Claude Code. Excellent tools, every one of them. World-class. But not Indian. Not sovereign.

And the irony struck me with real force. The AI was sovereign. The intelligence at the core was India’s own. But the tools — the assistants, the coding interfaces, the everyday instruments through which a developer actually touches that intelligence — those still came from elsewhere.

What good is a sovereign brain if you can only reach it through someone else’s body? That was the moment. I can point to it exactly. The gap was sitting right there on the screen, obvious and unfilled. A sovereign AI deserved a sovereign tool to work with it. Somebody should build that. And the voice from those YouTube videos came back, clear as anything: those who start building will thrive.

I decided to make it mine.

Sarvam-CLI I didn’t deliberate. I didn’t write a business plan. I didn’t ask anyone for permission — because by now I knew that building never required it.

I quickly created a Sarvam-CLI repo on GitHub and got started.

A command-line interface for India’s sovereign AI. A sovereign tool for a sovereign model. The missing piece on that integrations page, filled by someone who simply decided to fill it. Not a corporation. Not a funded team. One person who had spent two years turning fear into curiosity, curiosity into obsession, and obsession into the reflex of building.

I’m 51. I have cerebral palsy. I’m a Commerce graduate. I work in the Marketing team. I write pieces of Content (something that AI now does better and faster than me). There’s nothing Technical about whatever work I do.

And what I’ve done is right in front of you.

I certainly don’t write this for sympathy and I don’t write it for applause. I write it because somewhere there is a person reading this who has a reason — an age, a condition, a circumstance, a fear — that they believe disqualifies them from building. I want them to look at this repo and reconsider their fitness and their assumptions.

The level playing field the videos promised is real. I am the evidence. The tools that once demanded I overcome the limitations of my body before I could even express an idea — now meet me where I am. And I can see through them. That changes who gets to build. And in this era, the people who build are the people who thrive.

Sarvam-CLI — India’s 2nd Sovereign AI Coding Assistant.

India’s 2nd Sovereign AI Coding Assistant — well, that is my claim. To be very frank, I thought mine was the 1st. But when I did a Prior-Art search — I was proved wrong. A AI Startup in Bengaluru called CodeMate has the credit of developing India’s 1st Sovereign AI Coding Assistant. And, as of the time of writing this blog — CodeMate integrates with models from OpenAI, Google etc.. There’s no integration with models from Sarvam.ai (as of now).

It started with a small window in 2023, with a tool I refused to put down. It became 70 projects, a blog, and one moment of recognition on an integrations page. And it ended — for now — with a repo, a name, and a quiet, stubborn conviction that — the gap I saw was mine to close.

The playground was always open. I just walked in and Hit a 6.

If you’d like to follow along or contribute, the Sarvam-CLI repo is on GitHub. Builders welcome. Watchers welcome too — but maybe, after reading this, you’ll consider building instead…