As the appetite for AI continues to intensify, a new player has stepped into the spotlight: Deccan AI. Born only in October 2024, the young company has swiftly pulled in a staggering $25 million in a fresh Series A round, backed by A91 Partners and joined by heavyweight funds like Susquehanna International Group and Prosus Ventures. Their secret weapon? A dynamic team concentrated in Hyderabad, India, quietly powering the next generation of artificial intelligence from behind the scenes.
While big names like OpenAI and Anthropic hunker down with their own research, they often reach a crossroads when it comes to scaling models for the messy, unpredictable real world. Here, the most delicate and unforgiving work begins: sculpting and scrutinizing, vetting outputs, tightening every screw. It’s this vital, often-underappreciated domain—where reinforcement learning, meticulous data curation, and relentless model evaluation converge—that Deccan is determined to own.
The company’s service array runs wide—from honing a model’s prowess at writing code or becoming smarter digital agents, to connecting AI brains with an ever-widening array of external tools and APIs. Deccan doesn’t just sit in the passenger seat; it rolls up its sleeves with AI giants, helping them run rigorous evaluations, assemble judicious feedback, and build custom reinforcement learning environments. For clients in the enterprise world, there’s Helix, their precision-engineered evaluation suite, along with an automation platform aimed at wringing chaos from day-to-day operations.
But the AI landscape is shifting beneath everyone’s feet. What started as text-based wizardry is morphing into richer, more tactile “world models”—AI that isn’t just book smart, but can sense and act within the physical world. Think robotics. Vision systems. For Deccan, it’s not just about keeping pace; it’s about leading the charge.
Many of the world’s sharpest AI labs—among them Google DeepMind and Snowflake—have turned to Deccan’s expertise. At any given moment, the team is juggling roughly two dozen projects, serving around ten robust clients. The headcount? Roughly 125 full-timers, with a reservoir exceeding a million contributors—the kind who eat linear algebra for breakfast: grad students, seasoned PhDs, and, on busy months, up to ten thousand active minds in the mix.
The speed and precision demanded by this sector can’t be overstated. “Our customers’ tolerance for error is virtually zero,” says founder Rukesh Reddy. A wayward data tag or slipshod evaluation isn’t just a nuisance—it’s a live wire running right into deployed products. Time, too, is of the essence; requests for mountains of meticulously labeled data can surface late on a Friday and need filling come Monday morning.
Such pressure has stirred broader debate about gig labor, pay, and the human side of AI’s rapid expansion. At Deccan, compensation spans a staggering range: from $10 to as much as $700 per hour, with top contributors sometimes banking $7,000 a month. Talent, though, is disproportionately drawn from India—a calculated decision, Reddy says. Rather than scattering across dozens of markets, Deccan keeps the bulk of its workforce in one country, making quality control sharp and consistent. For especially niche projects—geospatial data, for example, or semiconductor design—the net widens to other countries, including the U.S.
This pattern—U.S. innovation, but Indian execution—mirrors a deeper reality in AI’s present order. India is fielding the world’s largest bench of technical talent, but for now, the hunt for next-level frontier models is largely confined to a handful of labs in America and China.
Still, Deccan’s not another face in the crowd. It didn’t crawl up from computer vision and classic data labeling; from birth, its ambitions lay in “GenAI”—meaning more sophisticated, nuanced tasks were the endgame from the outset. The gamble appears to be paying off: growth over the past year has been tenfold, with revenue now running into the double-digit millions (precise figures remain under wraps). Roughly four-fifths of that business comes from just five anchor clients: further proof that, in advanced AI, a handful of customers can move the whole market.
In a space awash with competitors—Meta-backed Scale AI, Surge AI, and others jockeying for dominance—Deccan’s sharp focus, centralized workforce, and insistence on quality may turn out to be its edge. The future of AI still needs its unseen architects. For now, Deccan is betting that India’s minds, coupled with discipline and speed, can outpace the rest.
