Deccan AI Raises $25M to Expand Post-Training AI Services with India-Based Experts
As the demand for training and refining artificial intelligence (AI) models accelerates, Deccan AI, a startup focused on post-training data and evaluation services, has successfully raised $25 million in its first major funding round. Much of this work is carried out by an India-based workforce of experts, underlining the country’s growing role in the global AI ecosystem.
The all-equity Series A round was led by A91 Partners, with additional participation from Susquehanna International Group and Prosus Ventures.
Meeting the Growing Need for Post-Training Support
While leading AI labs such as OpenAI and Anthropic develop core models internally, the post-training phase—covering data generation, evaluation, and reinforcement learning—is increasingly outsourced. This is driven by the need to make AI systems reliable for real-world applications. Deccan AI has emerged as a key player addressing this growing demand.
Founded in October 2024, Deccan offers services ranging from improving coding and agent capabilities in models to training systems to interact with external tools like application programming interfaces (APIs). The startup collaborates with frontier labs on generating expert feedback, conducting evaluations, and building reinforcement learning environments. It also serves enterprise clients through products such as its evaluation suite, Helix, and an operations automation platform.
As AI models expand beyond text into “world models” that understand physical environments, including robotics and vision systems, Deccan’s services are evolving alongside these technological trends.
A Growing Customer Base and Workforce
According to founder Rukesh Reddy, Deccan counts Google DeepMind and Snowflake among its clients. The company has onboarded about 10 customers and runs several dozen active projects simultaneously.
Headquartered in the San Francisco Bay Area with a large operations team in Hyderabad, Deccan employs around 125 people and maintains a network of over 1 million contributors, including students, domain experts, and PhDs. Typically, 5,000 to 10,000 contributors are active each month. About 10% of the contributor base holds advanced degrees, though the proportion is higher among those engaged in specialized projects, Reddy told TechCrunch.
A Competitive Market for AI Training Services
The AI training services sector has expanded rapidly alongside the rise of large language models. Competitors include Scale AI (owned by Meta), Surge AI, as well as startups Turing and Mercor, all offering data labeling, evaluation, and reinforcement learning support.
“Quality remains an unsolved problem,” Reddy said, emphasizing that tolerance for errors in post-training is “close to zero” since mistakes can directly impact production-level model performance. Post-training requires highly accurate, domain-specific data that is challenging to scale.
Reddy also noted that the work is time-sensitive, with labs often requiring large volumes of high-quality data within days, creating a constant tension between speed and accuracy.
Fair Compensation Amid Criticism of the Sector
The AI training industry has faced criticism regarding working conditions and pay, particularly due to the reliance on large pools of gig workers. Reddy stated that contributors on Deccan’s platform earn between $10 and $700 per hour, with top performers earning up to $7,000 a month.
India as a Hub for AI Talent
Despite serving primarily U.S.-based AI labs, most of Deccan’s contributors are located in India. While competitors such as Turing and Mercor source talent across multiple emerging markets, Deccan concentrates much of its workforce in India to maintain quality.
“Many of our competitors go to 100-plus countries to find the experts,” Reddy explained. “If you have operations in just one country, it becomes far easier to maintain quality.”
This strategy underscores India’s position in the global AI value chain—as a key supplier of talent and training data—while frontier model development remains dominated by a few U.S. and Chinese firms. Nevertheless, Deccan is beginning to source specialized expertise from other markets, including the U.S., for areas such as geospatial data and semiconductor design.
A “Born GenAI” Company
Unlike traditional data labeling firms that began with basic computer vision tasks, Deccan was built as a “born GenAI” company, focusing on high-skill work from inception. Over the past year, the startup has grown tenfold and now operates at a double-digit million-dollar revenue run rate, Reddy revealed, without disclosing exact figures. Around 80% of revenue comes from its top five customers, reflecting the concentrated nature of the frontier AI market.
0 Comments