Deccan AI Secures $25 Million to Enhance AI Model Training with India-Based Workforce
As demand surges for refining artificial intelligence models, Deccan AI – a startup specializing in post-training data and evaluation – has raised $25 million in its inaugural major funding round. A significant portion of this work is carried out by a skilled workforce based in India.
The all-equity Series A round was spearheaded by A91 Partners, with Susquehanna International Group and Prosus Ventures also participating in the investment, according to a report by TechCrunch.
The Rise of Outsourced AI Training
While leading AI labs like OpenAI and Anthropic develop core models internally, a growing trend sees the outsourcing of post-training tasks – including data generation, evaluation, and reinforcement learning – as companies strive for reliability in real-world applications. Deccan AI is emerging as a key player in this expanding market.
Founded in October 2024, Deccan AI offers services that range from improving coding and agent capabilities in AI models to training systems to interact with external tools via application programming interfaces (APIs). The company collaborates with frontier labs on tasks such as generating expert feedback, conducting evaluations, and building reinforcement learning environments. It also serves enterprises through products like its evaluation suite, Helix, and an operations automation platform.
The company’s work is evolving to encompass “world models” that enhance understanding of physical environments, including applications in robotics and vision systems. Deccan AI’s clientele includes Google DeepMind and Snowflake as reported by The Outpost.
India as a Hub for AI Talent
Deccan AI currently serves approximately 10 customers and manages around two dozen active projects. The startup, headquartered in the San Francisco Bay Area, has a substantial operations team in Hyderabad and employs around 125 people. It also leverages a network of over 1 million contributors, including students, domain experts, and PhDs, with 5,000 to 10,000 active contributors each month according to Tech in Asia.
Approximately 10% of Deccan’s contributor base holds advanced degrees, such as master’s and PhDs, with a higher proportion among active contributors depending on project requirements.
Addressing Quality and Accuracy in AI Training
The AI training services market has experienced rapid expansion alongside the growth of large language models. Companies like Scale AI and Surge AI, along with startups like Turing and Mercor, compete in providing data labeling, evaluation, and reinforcement learning services as noted by TechCrunch.
Deccan AI founder Rukesh Reddy emphasizes the critical importance of quality, stating that tolerance for errors in post-training is “close to zero” due to the potential impact on model performance. He highlights the complexity of post-training, requiring highly accurate, domain-specific data that is challenging to scale.
Reddy also points to the time-sensitive nature of the work, with AI labs often needing large volumes of high-quality data within days, creating a balance between speed, and accuracy.
Fair Compensation and a GenAI-Focused Approach
Deccan AI addresses concerns regarding working conditions and pay in the AI training sector by offering earnings ranging from $10 to $700 per hour, with top contributors earning up to $7,000 per month.
The company was designed as a “born GenAI” entity, focusing on higher-skill work from the outset, differentiating it from traditional data labeling firms that began with computer vision tasks. Deccan AI has experienced 10x growth over the past year and currently maintains a double-digit million-dollar revenue run rate. Approximately 80% of its revenue is generated from its top five customers, reflecting the concentrated nature of the frontier AI market.
Keep reading