Deccan AI, a rival to Mercor, secures $25M and taps Indian expert talent

As demand continues to rise for training and refining artificial intelligence models, Deccan AI — a company focused on post-training data and evaluation — has secured $25 million in its first significant funding round, with much of its operational work supported by a large expert workforce based in India.

The all-equity Series A round was led by A91 Partners, with participation from Susquehanna International Group and Prosus Ventures.

While leading AI labs such as OpenAI and Anthropic develop foundational models internally, a growing portion of post-training work — including data generation, evaluation, and reinforcement learning — is increasingly being outsourced. Deccan AI is positioning itself as part of a new wave of startups meeting this demand.

Established in October 2024, the company provides services to improve model capabilities, including enhancing coding performance, enabling agent-based workflows, and training systems to interact with external tools such as application programming interfaces (APIs).

The startup collaborates with frontier AI labs on tasks such as generating expert feedback, evaluating models, and creating reinforcement learning environments. It also serves enterprise clients through products such as its evaluation suite, Helix, and an operations automation platform. Its work is evolving as AI systems move beyond text-based capabilities into “world models” that better interpret physical environments, including robotics and vision-based systems.

Deccan’s customer base includes Google DeepMind and Snowflake. According to founder Rukesh Reddy, the company has onboarded around 10 clients and typically manages several dozen active projects at any given time.

Headquartered in the San Francisco Bay Area with a major operations team in Hyderabad, Deccan employs approximately 125 staff members. It also operates a large contributor network of more than one million individuals, including students, domain specialists, and PhD-level experts. Between 5,000 and 10,000 contributors are active in a typical month.

Reddy noted that roughly 10% of contributors hold advanced degrees, though this percentage can vary depending on project requirements, particularly among active contributors.

The broader market for AI training services has expanded rapidly alongside the rise of large language models. Companies such as Scale AI, its competitor Surge AI, and startups like Turing and Mercor are competing in areas including data labelling, evaluation, and reinforcement learning.

Reddy emphasised that quality remains a major challenge in this space, noting that tolerance for errors in post-training work is extremely low because inaccuracies can directly impact model performance in real-world deployment. This makes post-training significantly more complex than earlier stages, requiring precise, domain-specific data that is harder to scale efficiently.

He also pointed out that the work is highly time-sensitive, with AI labs often requiring large volumes of high-quality data within tight deadlines, making it difficult to balance speed and accuracy.

The sector has faced scrutiny over labour conditions and compensation, particularly due to reliance on gig workers for training data generation. Reddy said contributors on Deccan’s platform earn between $10 and $700 per hour, with top performers earning up to $7,000 per month.

India emerges as a key hub for AI training talent.

Although Deccan primarily serves U.S.-based AI labs, the majority of its contributors are located in India. Competitors such as Turing and Mercor also recruit talent from the region, though they operate across a wider range of countries.

Deccan has chosen to concentrate much of its workforce in India to maintain better control over quality. Reddy explained that limiting operations to a single country simplifies management compared to sourcing talent from dozens of regions globally.

This strategy highlights India’s role in the global AI ecosystem as a major supplier of skilled labour and training data, rather than a primary developer of frontier models, which remain largely concentrated among a handful of U.S. firms and select companies in China.

At the same time, Deccan has begun expanding its talent sourcing to other regions, including the United States, particularly for specialised expertise in aeronautical data and semiconductor design.

Reddy described Deccan as a “born GenAI” company, distinguishing it from traditional data-labelling firms that originated in computer vision. From the outset, the company has focused on higher-skill, more complex work.

Over the past year, Deccan has experienced rapid growth, increasing its scale tenfold and reaching a double-digit million-dollar annual revenue run rate, though specific figures were not disclosed. Approximately 80% of its revenue comes from its top five customers, reflecting the highly concentrated nature of the frontier AI market.