Mercor Competitor Deccan AI Raises $25M, Sources Experts From India

Deccan AI Raises $25M to Fix the Problem in AI Training 

As frontier labs race to build smarter models, a Hyderabad-rooted startup is quietly solving the quality crisis that no one talks about — and it just landed $25 million to scale faster. 

Credit: Deccan AI

Deccan AI, a startup specializing in AI post-training data and model evaluation, has raised in its first major funding round. The Series A was led by A91 Partners, with participation from Susquehanna International Group and Prosus Ventures. Founded just in October 2024, the company has already landed clients such as Google DeepMind and Snowflake — and it's doing it with a workforce rooted in India. 

Why AI Post-Training Is the Industry's Hardest Unsolved Problem 

Building an AI model is only half the battle. The harder, costlier, and less glamorous work happens after — refining how the model behaves, evaluating its outputs for errors, and teaching it to interact reliably with real-world software. This phase, known as post-training, is where most AI systems either succeed or quietly fall apart. Deccan AI sits squarely at this pressure point, offering services ranging from expert feedback generation to reinforcement learning environments and evaluation tooling.

Founder Rukesh Reddy is direct about the stakes: tolerance for errors in post-training is "close to zero," because a single mistake at this stage can ripple into broken behavior in production. That makes the work fundamentally different — and harder — than earlier stages of model development, demanding domain-specific expertise and near-perfect accuracy at speed.

"Quality remains an unsolved problem."

India at the Center of a Global AI Supply Chain

While frontier AI models are built by a handful of U.S. companies and a few players in China, the expertise powering those models increasingly flows from India. Deccan AI has leaned into this deliberately, concentrating its contributor network in one country rather than spreading across dozens of markets. The reasoning is strategic: tighter geographic focus makes quality control significantly more manageable. The company currently draws on a pool of over one million contributors, with five thousand to ten thousand active in any given month.

Approximately ten percent of contributors hold advanced degrees — master's or PhDs — though that share is much higher among active contributors on technically demanding projects. Top earners on the platform take home up to seven thousand dollars a month, with hourly rates ranging from ten to seven hundred dollars depending on the complexity of the task. 

From Coding Help to Robotics: How Deccan AI Is Expanding 

Deccan's work isn't limited to text-based AI. As models evolve toward understanding physical environments — what the industry calls "world models" — Deccan is positioning itself to support robotics and vision systems as well. The company's product lineup includes Helix, an evaluation suite, and an operations automation platform aimed at enterprise clients. About ten customers are currently onboarded, running a couple of dozen active projects at any given time. 

Revenue is growing fast. The company has achieved a double-digit million-dollar annual run rate after growing ten times over the past twelve months. That said, the business remains concentrated: roughly eighty percent of revenue comes from its top five clients, a reality that mirrors the narrow structure of the frontier AI market itself. 

A "Born GenAI" Company Built for the Moment 

Reddy describes Deccan as a "born GenAI" company — not a legacy data labeling firm that retrofitted itself for the AI era. That distinction matters. Traditional data annotation businesses built their foundations on computer vision tasks; Deccan was designed from day one for higher-skill, language-focused work. With roughly 125 full-time employees split between the San Francisco Bay Area and Hyderabad, the company blends Silicon Valley proximity with Indian talent depth. 

The startup is also beginning to look beyond India for niche expertise. Specialists in geospatial data and semiconductor design are now being sourced from a handful of other markets, including the United States — a signal that Deccan's ambitions are quietly expanding even as its core identity remains India-first. 

In a market crowded with data labeling providers, Deccan AI is betting that quality over scale is the winning formula. With $25 million in fresh funding and clients at the top of the AI hierarchy already signed on, it's a bet that looks increasingly well-placed.