I Didn't Start a Company. I Just Kept Building Until I Had One.

Nobody has a clean founder story. Not really.

The ones you read in interviews are edited. The failures are framed as character building. The pivots sound strategic. The timing looks intentional.

Mine isn't like that. So I'm going to tell it straight.

It started in 10th grade

I was 15. I built something called ShoppingDNA. The idea was a product comparison tool for Indian e-commerce. Users search for a product, see prices across platforms, pick the best one.

I thought it was a good idea. I spent months building it. I got it working. I showed it to people.

It ran for 8 months. Then it stopped.

What killed it: I built something that solved a problem I had personally, but I had no plan for getting users. Zero. I assumed if the product was good, people would show up. They didn't show up. I had no money to run ads. I had no audience. I had no network that could give it any real distribution.

I was 16 when it ended. Disappointed but not destroyed. I kept coding.

The years in between

After ShoppingDNA I went through a stretch of building smaller things. Freelance web work. College projects. A few tools I used myself but never turned into products.

I won a FinTech contest at Techfest 2020. That felt good. But it was a contest, not a business. The distinction matters.

During college I got involved in Rotaract, became International Service Director. I was a Microsoft Learn Student Ambassador. I was learning how to operate in communities, how to lead things, how to talk to people who were further ahead than me. Skills I didn't know I was accumulating.

What I was actually doing in these years: getting better at things that would later matter. I just didn't know what I was preparing for yet.

2021 — GPT-3 changes everything

I was 19 or 20. GPT-3 had come out. I got early access. I started building with it immediately.

I wasn't thinking about a company yet. I was just genuinely excited. This was a different kind of technology. You could give it instructions and it would generate text that actually made sense. The applications were everywhere if you knew where to look.

I started working with a project out of Melbourne called Polymorphix. Early LLM work. Prompt engineering before that phrase existed in any serious way. I was figuring out what these models could actually do when you pushed them.

That period taught me something important: most people were using these models the same way you'd use a search engine. They weren't thinking about them as building blocks for systems. I was. That gap between how most people used AI and what it could actually do — that's where CyberCore would eventually live.

Enterprise India Fellowship

I joined the Enterprise India Fellowship as a Technology Catalyst. This is where things got serious.

I was working with real companies, real workflows, real problems. Not toy projects. Not hackathon demos. Actual operations that people depended on. CRM systems. Automation pipelines. Processes that broke if something didn't work.

I automated 22+ business processes across that time. Brought manual effort down by 30-50% in the workflows I touched. These weren't impressive numbers to put on a LinkedIn profile. They were just the results of sitting with a problem long enough to actually fix it.

This is also where I learned that AI implementation is mostly not an AI problem. It's a process problem. It's a people problem. The model is usually fine. The hard part is understanding what a business actually does at a granular level, and then building something that fits into how it runs rather than forcing it to change.

Most AI projects fail not because the model is bad. They fail because someone automated the wrong thing.

When CyberCore became real

I can't give you a precise moment. There was no day I decided to start a company.

What happened instead: I kept getting asked to do AI work. Not small stuff. Founders calling me, asking if I could help them build AI into their operations. Product teams who had tried the obvious tools and hit a wall. Companies that had bought an AI platform, connected it, and then found that it didn't actually do what they needed.

At some point I noticed I was doing company work without calling it a company. CyberCore became the name for what I was already doing.

The positioning came from something I kept seeing: most AI builds were chatbots. Question-answer systems. Things that made information slightly more accessible. What was missing was AI that actually finished tasks. End to end. Agentic systems that could take a workflow and run it without someone babysitting it.

That's what CyberCore focuses on. Not chatbots. AI systems that do the work.

What I know now that I didn't know at 15

ShoppingDNA failed because I built before I understood the market. I thought a good product was enough. It's not. You need a way to reach the people who need it. And you need to understand what they actually need, not just what you think they need.

Building CyberCore is different because I spent years inside real companies, watching real problems. I didn't start with a product idea. I started with an understanding of where AI systems actually break down, what businesses are trying to do with them, and what they're not getting from the current solutions.

The 10 years in between weren't a detour. They were the research.

What CyberCore looks like now

We build agentic AI systems for startups and small enterprises. Python, LangChain, LangGraph, LlamaIndex, vector databases, RAG pipelines. The whole stack.

Our clients are typically founders or CTOs at 10 to 50 person companies. They're building AI products or trying to automate operations. They don't have a dedicated AI hire yet. They need someone who can sit with their actual workflows and build something that works reliably, not a demo that looks good for five minutes.

I'm 24. The company is young. We have a long way to go.

But I've been building since I was 15. That part isn't new.