In a world obsessed with Silicon Valley giants, Rohan Mehta is quietly building a global empire from Pune. As the founder of Cognic, an enterprise AI SaaS that automates complex data engineering pipelines, Rohan has shown that deep tech innovation is thriving in India. Bootstrapped to $5M ARR before taking a single dime of institutional capital, Cognic is reshaping how Fortune 500 companies process data.
We discussed the reality of building a highly technical product outside the valley bubble, managing remote talent, and the future of agentic AI.
Prior to Cognic, I worked as a lead data architect in New York. I realized that my team, comprised of brilliant PhDs, was spending 70% of their time just cleaning and structuring data. The actual machine learning part was an afterthought because the plumbing was so broken. I saw an opportunity to let AI do the heavy lifting of data normalization so engineers could get back to building models.
The main con used to be capital access, but that's largely resolved in 2026. The pros, however, are immense. Our burn rate is a fraction of a Valley startup, allowing us to be uncharacteristically patient with our product roadmap. We didn't have to launch half-baked features to secure the next funding round. Plus, the engineering talent in India is phenomenal. We have builders who are incredibly resilient and loyal to the vision.
Absolutely deliberate. When you take early money, you're not just selling equity; you're selling the option to pivot freely. We were dealing with a complex B2B product, and I knew it would take 18-24 months just to get the architecture right. I didn't want a boardroom pressuring me to increase marketing spend when the product wasn't ready to scale. We raised our Series A only when we needed a global sales machinery.
A thin wrapper around a foundational model is not a moat. That gets crushed the moment OpenAI or Anthropic releases an API update. True defensibility comes from proprietary workflows, deep integrations into legacy enterprise systems, and compounding domain knowledge. We don't just generate code; we securely execute, test, and maintain data pipelines entirely on-premise for banks. The trust and integration depth is our moat, not the LLM.
About a year in, we built a beautiful dashboard that clients specifically asked for. We spent three months perfecting it. When we launched, usage was zero. It turned out clients wanted the data automated via API; they never actually wanted to log into another dashboard. It taught me a brutal lesson: listen to a customer's pain, but never blindly build their proposed solution.
The first two years were incredibly lonely. My own friends and even some family members thought I was wasting my US-earned savings on a 'research project' that sounded too complex to ever work. I didn't have a mentor, and the local ecosystem back then was more focused on services than deep SaaS products. What saved me was the internet. I found my tribe in obscure Discord servers and open-source communities where people cared about the tech, not the geography. I learned to build in public before it was a trend. When the people around you don't support you, you have to build your own support system online. That isolation actually forced me to focus entirely on the code, which is why Cognic's core architecture is so robust today.
The fear of failing the people who took a chance on me. When you have employees who’ve left stable jobs to join your 'crazy team', your responsibility shifts from yourself to them. There were months when I didn't take a salary just to keep the lights on. What kept me going was a simple realization: the struggle is the price of admission for anything extraordinary. If it were easy, it wouldn't be worth building.