Deccan AI Secures $25 Million to Enhance AI Model Training Services
In response to the increasing demand for sophisticated AI model training and refinement, Deccan AI, a startup focused on post-training data and evaluation services, has successfully raised $25 million in its inaugural major funding round. A significant portion of this work is supported by a skilled workforce based in India.
The Series A funding round was entirely equity-based and led by A91 Partners, with notable contributions from Susquehanna International Group and Prosus Ventures.
Outsourcing Post-Training Work for Enhanced Reliability
While leading AI labs like OpenAI and Anthropic develop core models internally, the outsourcing of post-training tasks—ranging from data generation to evaluation and reinforcement learning—has become increasingly common. As companies strive to ensure the reliability of AI systems in practical applications, Deccan is positioning itself as a crucial player in this emerging market.
Comprehensive Services for AI Model Improvement
Founded in October 2024, Deccan specializes in a broad spectrum of services designed to enhance the performance of AI models. These include assistance in improving coding and agent capabilities as well as training models for seamless interaction with external tools, such as application programming interfaces (APIs) that integrate AI systems with software applications.
Deccan collaborates with leading AI labs on various critical tasks, including generating expert feedback, conducting evaluations, and creating reinforcement learning environments. Additionally, it caters to enterprises with products like its evaluation suite, Helix, and an operations automation platform. The firm’s focus is evolving as AI models transition from text-based capabilities to more complex “world models,” which have a better grasp of physical environments, robotics, and vision systems.
Expanding Client Portfolio and Project Pipeline
According to founder Rukesh Reddy, Deccan’s client roster includes well-known entities such as Google DeepMind and Snowflake. The startup has onboarded approximately 10 customers and typically manages a couple of dozen active projects. The company, headquartered in the San Francisco Bay Area with a substantial operational team in Hyderabad, employs around 125 individuals, supported by a network of over 1 million contributors, including students, domain specialists, and PhDs. Reddy noted that 5,000 to 10,000 of these contributors are actively engaged each month.
Quality Control and Industry Growth
Reddy emphasized the importance of quality in post-training processes, stating, “Quality remains an unsolved problem.” With error tolerance in post-training work being extremely low—given that mistakes can adversely affect model performance in real-world applications—this phase requires precise, domain-specific data that is often challenging to scale. The time-sensitive nature of AI projects further complicates this landscape, with labs frequently seeking large quantities of high-quality data on tight deadlines.
India Emerges as a Key Player in AI Training Talent
Despite serving primarily U.S.-based AI labs, a majority of Deccan’s contributors are sourced from India. Similar to competitors like Turing and Mercor, which also draw talent from India, Deccan’s focus is geared towards maintaining quality by limiting its operational reach. Reddy stated that managing a workforce from a single country simplifies quality assurance compared to a broader geographical strategy.
This operational model underscores India’s growing significance in the global AI value chain, positioning the country as a crucial supplier of talent and training data, while leading AI model development remains largely concentrated within a few U.S. corporations and select firms in China. However, Reddy mentioned that Deccan is beginning to tap into other markets, including the U.S., to find specialists with niche skills in fields such as geospatial data and semiconductor design.
Deccan AI’s Rapid Growth Trajectory
Positioned as a “born GenAI” entity, Deccan differentiates itself from traditional data labeling companies, focusing on higher-skill tasks from the outset. Reddy reported a remarkable tenfold growth over the past year, with the company now achieving a revenue run rate in the double-digit million-dollar range, although he opted not to disclose specific figures. Currently, around 80% of Deccan’s revenue is generated from its top five clients, illustrating the concentrated nature of the frontier AI market.