Deccan AI, a burgeoning startup specializing in post-training data and evaluation services for artificial intelligence models, has successfully secured $25 million in its inaugural major funding round. This significant investment comes as demand for advanced AI model training and refinement continues to surge, with a substantial portion of Deccan AI's operations powered by its expert workforce based in India.
The Series A funding round, conducted entirely through equity, was spearheaded by A91 Partners. Notable participation also came from Susquehanna International Group and Prosus Ventures, underscoring strong investor confidence in Deccan AI's mission.
While leading AI research institutions such as OpenAI and Anthropic develop their foundational models internally, the intricate and labor-intensive post-training phase—encompassing data generation, rigorous evaluation, and reinforcement learning—is increasingly being outsourced. This trend is driven by companies' imperative to ensure their AI systems are robust and reliable for real-world applications. Deccan AI is rapidly establishing itself as a key player among the new generation of startups catering to this escalating demand.
Established in October 2024, Deccan AI offers a comprehensive suite of services. These range from enhancing AI models' coding and agent capabilities to training systems for seamless interaction with external tools, including Application Programming Interfaces (APIs) that bridge AI models with various software systems.
The startup collaborates with frontier AI laboratories on critical tasks such as eliciting expert feedback, conducting thorough evaluations, and constructing sophisticated reinforcement learning environments. Concurrently, Deccan AI supports enterprises with its proprietary products, including the Helix evaluation suite and an operations automation platform. The scope of its work is also expanding beyond text-based models, moving into "world models" that aim for a deeper understanding of physical environments, encompassing applications in robotics and advanced vision systems.
According to the company, Deccan AI boasts an impressive client roster that includes industry giants like Google DeepMind and Snowflake. Founder Rukesh Reddy revealed in an interview that the company has onboarded approximately 10 customers and manages several dozen active projects concurrently.
Headquartered in the San Francisco Bay Area, with a significant operations team situated in Hyderabad, Deccan AI employs around 125 individuals. The company also leverages an expansive network of over 1 million contributors, comprising students, domain experts, and PhDs. Reddy informed TechCrunch that typically between 5,000 and 10,000 contributors are actively engaged in projects each month.
Reddy further clarified that approximately 10% of Deccan AI's overall contributor base holds advanced degrees, such as master's and PhDs. However, this proportion tends to be higher among active contributors, varying based on specific project requirements.
The market for AI training services has witnessed explosive growth in parallel with the proliferation of large language models. Companies like Meta-owned Scale AI, its competitor Surge AI, and emerging startups such as Turing and Mercor are all vying to provide essential services including data labeling, evaluation, and reinforcement learning.
“Quality remains an unsolved problem,” stated Reddy, emphasizing that the tolerance for errors in post-training processes is "close to zero." This stringent requirement stems from the fact that mistakes at this stage can directly and significantly impair a model's performance in production. Consequently, post-training becomes far more complex than earlier development stages, demanding exceptionally accurate, domain-specific data that is inherently challenging to scale.
The work is also characterized by its extreme time sensitivity, Reddy added. AI labs frequently require substantial volumes of high-quality data within a mere matter of days, presenting a formidable challenge in balancing the imperative for speed with uncompromising accuracy.
Historically, the sector has faced considerable criticism regarding working conditions and compensation, often relying on large pools of gig workers for training data generation. Reddy addressed this by stating that earnings on Deccan AI’s platform typically range from $10 to $700 per hour, with top contributors having the potential to earn up to $7,000 monthly.
Despite its customer base being predominantly U.S.-based AI laboratories, the majority of Deccan AI's contributors reside in India. While competitors such as Turing and Mercor also source contractors from India, they generally extend their operations across a broader spectrum of emerging markets.
Deccan AI's strategic decision to concentrate a significant portion of its workforce in India was driven by a commitment to superior quality management, Reddy explained. He noted, “Many of our competitors go to 100-plus countries to find the experts. If you have operations in just one country, it becomes far easier to maintain quality.”
This concentrated approach underscores India's current pivotal role within the global AI value chain—primarily as a supplier of talent and training data. This stands in contrast to the development of frontier AI models, which largely remains concentrated among a select group of U.S. companies and a few key players in China.
However, Reddy clarified that Deccan AI has recently begun to source specialized talent from a few other markets, including the U.S., particularly for niche expertise in areas such as geospatial data and semiconductor design.
Reddy highlighted that Deccan AI was purposefully built as a “born GenAI” company, distinguishing it from traditional data labeling firms that typically originated with computer vision tasks. This foundational approach has enabled Deccan AI to focus on higher-skill, more complex work from its very inception.
Deccan AI has experienced a tenfold growth over the past year and is currently operating at a double-digit million-dollar revenue run rate, though Reddy refrained from disclosing specific figures. He further noted that approximately 80% of the company's revenue is generated from its top five customers, a statistic that reflects the concentrated nature of the frontier AI market.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.