The story behind Koovis AI.
Koovis AI exists because of a simple frustration: the gap between what AI can do in a demo and what it actually does in production is enormous. After years of watching promising AI projects die in the “last mile” — the monitoring, the failover, the operational discipline — we decided to build products that cross that gap ourselves.
We're a product company, not an agency. We build our own AI products, use them daily, and iterate based on what we learn from real usage.
Four products, one engine. Koovis Workforce (autonomous AI workforce for founders) — MIT-licensed engine running in production since March 2026. Koovis Pulse (Indian-language AI UGC ads for D2C brands). WealthPilot Research (ablation-first research on Indian listed companies, 5-year OOS validated). Koovis Studios (AI pre-visualization for Telugu cinema and beyond). Different domains, same philosophy — build things that actually work when nobody's watching.
What we're building toward
We want to build AI products that people quietly rely on every day. Not the flashiest tools in someone's demo reel. Not the most hyped product on launch day. The ones that just work — reliably, consistently — and earn trust over time through performance, not promises.
Rajesh Kolachana
Founder & CEO
The path here wasn't straight
I started in structural engineering — IIT Roorkee for my B.Tech, then IISc Bangalore for my M.Tech. Scored GATE AIR 5 (top 0.013% nationally), which in Indian engineering circles opens every door. But somewhere between finite element analysis and optimization theory, I realized the same mathematical frameworks that model bridges and buildings could model human behavior, markets, and decisions. That pivot changed everything.
My first real data science role was at InMobi, where I won the Rising Star Award and scaled an ad account from $3K to $80K in daily spend. Then AgreeYa Solutions, building pricing optimization models for Best Buy, Sam's Club, and Dick's Sporting Goods. Both taught me what production ML actually looks like — messy data, tight deadlines, and systems that have to work at 2 AM on a Saturday.
Seven years at Amazon
Amazon is where I learned what it means to build at scale. Seven years as a Senior Data Scientist, shipping ML systems across global marketplaces — recommendation engines, NLP-driven review ranking, Bayesian reorder models, paid- advertising optimization, and an NL-to-SQL tool that went from hackathon project to a production tool used by thousands of account managers worldwide.
The real takeaway wasn't the scale — it was the discipline. The operational rigor. The understanding that a model is maybe 20% of a production ML system; the rest is pipelines, monitoring, failover, and the boring engineering that keeps things running at 3 AM.
Why I left
After 11 years in the industry — building systems for other people's products — I wanted to build my own. To apply everything I learned at Amazon scale, use it myself every day, and see if one person with the right tools and the right discipline can ship things that matter.
Koovis is intentionally small. Solo for now; when I hire, it's because a specific constraint demands it, not because growth metrics say so. I publish negative findings alongside positive ones. I open-source our core engine because the best infrastructure compounds when others can run it too.
Koovis AI is the bet. It's early, and there's a lot left to build. But the Workforce engine is running in production, WealthPilot's 5-year OOS ablation paper is in submission, Pulse is taking its first customer pilots, and Studios is shipping Demo v2. Ask me again in a year.
What we believe.
Ship, Don’t Demo
A working product in production teaches you more in a week than a prototype teaches you in a year. We ship first, polish second, and learn from every deployment.
Honest Architecture
Not every problem needs ML. Sometimes the best solution is a well-crafted SQL query. We pick the simplest tool that actually solves the problem — ego aside.
Own the Outcome
Every product we build is something we use ourselves, every day. That’s the bar. If it doesn’t make our own lives better, it’s not ready for anyone else.
Build in Public
We share what works, what breaks, and what we learn along the way. It’s slower than slick marketing, but it builds the kind of trust you can’t buy.
See what we're building.
Four products, one engine. Pick the one that fits your work.