Deccan AI Raises $25 Million and Bets on India to Dominate AI

Deccan AI: A Strategic Fundraising for the Future of AI

In response to the growing demand for training and refining artificial intelligence models, Deccan AI, a startup specializing in post-training data and evaluation, has recently raised $25 million. This funding, crucial for its expansion, was secured during its first major funding round. A significant part of its work relies on a team of experts based in India, a strategic choice that reflects the increasing importance of this country in the field of AI.

The Series A funding round, entirely equity-based, was led by A91 Partners. Other notable investors include Susquehanna International Group and Prosus Ventures. This capital injection aims to enhance Deccan AI's ability to provide critical services in the AI domain.

Outsourcing: A Key Strategy for AI Labs

While leading AI labs such as OpenAI and Anthropic focus on the internal development of foundational models, much of the post-training work is now outsourced. This process includes data generation, evaluation, and reinforcement learning, essential steps to ensure the reliability of systems in real-world conditions. Deccan AI positions itself as an emerging player responding to this growing demand.

Founded in October 2024, Deccan AI offers a range of services from enhancing coding capabilities to training systems to interact with external tools, such as application programming interfaces (APIs). These services connect AI models to software systems, thereby enhancing their utility and efficiency.

Collaboration with Industry Giants

The startup collaborates with leading labs for tasks such as generating expert feedback, conducting evaluations, and creating reinforcement learning environments. It also offers products like its evaluation suite, Helix, and an operations automation platform. Furthermore, its work is evolving to include models that better understand physical environments, particularly in robotics and vision systems.

Among its prestigious clients, Deccan AI counts Google DeepMind and Snowflake. The startup has integrated around 10 clients and manages several dozen active projects at any given time. According to its founder, Rukesh Reddy, this momentum reflects the trust placed in Deccan by industry leaders.

A Diverse and Skilled Workforce

Based in the San Francisco Bay Area, Deccan AI has a significant operational team in Hyderabad. The company employs around 125 people and relies on a vast network of over one million contributors, including students, domain experts, and PhD holders. Each month, between 5,000 and 10,000 contributors are active, bringing their expertise to meet project requirements.

About 10% of Deccan's contributors hold advanced degrees, such as master's and doctoral degrees. This proportion is even higher among active contributors, depending on the specific needs of projects, Reddy emphasizes.

A Rapidly Expanding Market

The market for AI training services is experiencing rapid growth, driven by the rise of large language models. Companies like Scale AI, owned by Meta, and its competitor Surge AI, along with startups such as Turing and Mercor, are competing to provide data labeling, evaluation, and reinforcement learning services.

Rukesh Reddy points out that quality remains a major challenge, with an almost zero tolerance for errors in post-training. Mistakes can indeed have a direct impact on the performance of models in production, making this stage more complex than previous phases. It requires domain-specific, precise data that is difficult to scale.

The Importance of Speed and Accuracy

Work in this sector is also highly time-sensitive. AI labs sometimes need large volumes of high-quality data in just a few days, complicating the balance between speed and accuracy.

This sector has faced criticism regarding working conditions and compensation, particularly due to the use of large groups of gig workers to generate training data. Reddy notes that earnings on the Deccan platform range from about $10 to $700 per hour, with top contributors able to earn up to $7,000 per month.

India: An Expanding Talent Hub

Although its clients are primarily AI labs based in the United States, the majority of Deccan's contributors are located in India. Competitors like Turing and Mercor also recruit contractors in this country but operate across a broader range of emerging markets.

Deccan has chosen to concentrate a large part of its workforce in India to better manage quality, Reddy explains. "Many of our competitors go to over 100 countries to find experts," he says. "If you have operations in just one country, it becomes much easier to maintain quality."

This approach highlights India's current position in the global AI value chain, as a provider of talent and training data, rather than as a developer of cutting-edge models, which remain concentrated among a few American companies and some players in China.

Targeted International Expansion

However, Reddy mentions that Deccan has begun to seek talent in a few other markets, including the United States, for niche expertise in geospatial data and semiconductor design.

Deccan has been built as a "GenAI-born" company, unlike traditional data labeling companies that started with computer vision tasks. This means it has focused on higher-skill work from the outset.

Over the past year, Deccan has experienced tenfold growth and now reports annual revenue in the double digits in millions of dollars, according to Reddy, who did not wish to share specific details. About 80% of its revenue comes from its top five clients, reflecting the concentrated nature of the cutting-edge AI market.