Deccan AI, a startup that provides post-training data and evaluation work to meet the growing demand for training and improving AI models, has raised $25 million in its first major funding round, with much of the work performed by a workforce of experts based in India.
The all-stock Series A round was led by A91 Partners, with participation from Susquehanna International Group and Prosus Ventures.
While frontier AI labs like OpenAI and Anthropic build core models in-house, much of the post-training work, from data generation to evaluation to reinforcement learning, is increasingly outsourced as companies seek to increase the reliability of their systems in real-world use. Deccan is emerging as one of the new startups to meet that demand.
Founded in October 2024, Deccan offers services ranging from helping to code models and improve agent capabilities to training systems for interacting with external tools, such as application programming interfaces (APIs) that connect AI models to software systems.
The startup works with Frontier Labs on tasks such as generating expert feedback, performing evaluations, and building reinforcement learning environments, while serving enterprises through products such as its evaluation suite, Helix, and operational automation platform. Research is also evolving as models move beyond text to so-called “world models” that better understand the physical environment, such as robotics or visual systems.
The company says Deccan’s customers include Google DeepMind and Snowflake. Founder Lukesh Reddy (pictured above) said in an interview that the company employs about 10 customers and has dozens of active projects running at any given time.
The startup is headquartered in the San Francisco Bay Area with a large operations team in Hyderabad, employs approximately 125 people, and relies on a network of more than 1 million contributors, including students, professionals, and PhDs. Reddy told TechCrunch that it typically has about 5,000 to 10,000 contributors per month.
tech crunch event
San Francisco, California
|
October 13-15, 2026
About 10% of Deccan’s contributor base has advanced degrees, such as master’s degrees or PhDs, but Reddy said the percentage can be higher among active contributors depending on the requirements of the project.
The market for AI training services is rapidly expanding with the rise of large-scale language models, with companies like Meta-owned Scale AI and its rival Surge AI, as well as startups Turing and Mercor, competing to provide data labeling, evaluation, and reinforcement learning services.
“Quality is still an open issue,” Reddy said, adding that the tolerance for errors after training is “close to zero” as mistakes can directly impact the model’s performance in production. This makes the post-training task more complex than the initial stage and requires highly accurate domain-specific data that is difficult to scale.
He said the work is very time-sensitive, and AI labs can require large amounts of high-quality data within days, making it difficult to balance speed and accuracy.
The sector, which often uses large numbers of gig workers to generate training data, has faced criticism over working conditions and pay. Earnings on Deccan’s platform range from about $10 to $700 an hour, with top contributors earning up to $7,000 a month, Reddy said.
India emerges as a hub for AI talent development
While the majority of Deccan’s customers are US-based AI labs, most of Deccan’s contributors are based in India. Competitors such as Turing and Melkor also source contractors domestically, but operate in a broader range of emerging markets.
Deccan has chosen to centralize much of its workforce in India to improve quality control, Reddy said. “Many of our competitors travel to more than 100 countries to find experts,” he says. “If you operate in just one country, it’s much easier to maintain quality.”
This approach highlights India’s current position in the global AI value chain: as a supplier of talent and training data rather than a developer of frontier models. The frontier model remains concentrated in a small number of US companies and a small number of Chinese companies.
But Reddy said Deccan has begun sourcing talent from several other markets, including the United States, for niche expertise in geospatial data and semiconductor design.
Reddy said Deccan was founded as a “born GenAI” company, as opposed to traditional data label companies that started with computer vision work. This means the company has focused on higher-skill jobs from the beginning.
Deccan has grown 10 times over the past year and now has double-digit million dollar revenue, Reddy said, without providing further details. About 80% of the company’s revenue comes from its top five customers, reflecting the concentrated nature of the frontier AI market, he added.
