As demand continues to grow for training and refining AI models, Deccan AI [https://www.deccan.ai/] has managed to raise $25 million in its first major funding round. This investment will support the company as it supplies post-training data and evaluation services. A substantial portion of this work is completed by a highly skilled workforce based in India.

The recent all-equity Series A round was led by A91 Partners and also included Susquehanna International Group and Prosus Ventures as participants. While leading AI labs such as OpenAI and Anthropic typically develop core models internally, they increasingly outsource post-training tasks—ranging from data generation to evaluation and reinforcement learning as they aim to make systems reliable in real-world scenarios. Deccan is emerging among a new generation of startups meeting this demand.

The company, founded in October 2024, supports projects that enhance various model capabilities, from coding to interacting with external tools like APIs—which link AI to broader software systems. Typically, Deccan works with labs on expert feedback generation, evaluations, and reinforcement learning environments. Simultaneously, it services enterprise clients with products like the Helix evaluation suite and an operations automation platform. The evolving nature of AI also means Deccan’s work is starting to include more complex “world models” for robotics and vision systems.

Deccan’s Customer Base, Workforce, and Market Position

Deccan counts major names like Google DeepMind and Snowflake among its customers. According to founder Rukesh Reddy, the company has around 10 customers and a couple dozen active projects at any time. Its headquarters are in the San Francisco Bay Area, but a large operations team based in Hyderabad drives its efforts. Currently, the startup employs about 125 people and manages a contributor network of over 1 million, of whom 5,000 to 10,000 are typically active each month. This diverse group includes students, domain experts, and PhDs.

Notably, around 10% of Deccan’s contributors hold advanced degrees, with this share increasing for more specialized projects. The broader AI training service market has been expanding rapidly [https://www.ft.com/content/0cab0fcd-e355-40e8-83a3-2ad5066d7b48?syn-25a6b1a6=1], in step with the rise of large language models. Startups like Meta-owned Scale AI [https://techcrunch.com/2025/06/13/scale-ai-confirms-significant-investment-from-meta-says-ceo-alexandr-wang-is-leaving/], competitor Surge AI [https://www.bloomberg.com/news/articles/2025-07-30/scale-rival-surge-ai-in-talks-for-funding-at-25-billion-value], and others such as Turing [https://techcrunch.com/2025/03/06/turing-a-key-coding-provider-for-openai-and-other-llm-producers-raises-111m-at-a-2-2b-valuation/] and Mercor [https://techcrunch.com/2025/09/09/sources-ai-training-startup-mercor-eyes-10b-valuation-on-450m-run-rate/] compete fiercely in areas like data labeling, evaluation, and reinforcement learning.

Founder Rukesh Reddy highlights that quality remains a challenge in post-training. Tolerance for error is minimal because mistakes have direct consequences for production performance. This complexity sets post-training apart, as it demands highly accurate, domain-specific data that is hard to scale. The work is also time-sensitive. Many labs require large volumes of top-quality data within days, which complicates the balance between speed and accuracy.

India’s Role in the Global AI Talent Chain

Working conditions and pay have occasionally sparked criticism in the sector [https://labourreview.org/the-human-cost-of-training-ai/]. Although tasks are performed by a broad contingent of gig workers, Deccan’s pay scales—from $10 up to $700 per hour, with top contributors earning as much as $7,000 monthly—suggest notable earning potential.

While most Deccan clients are U.S.-based AI labs, the majority of its workforce operates from India. Competitors like Turing and Mercor also recruit there [https://the-ken.com/story/train-ai-models-earn-up-to-50-hr-these-firms-turn-to-indian-phds-techies-actors/], but engage contributors from many emerging markets [https://www.bloomberg.com/news/newsletters/2025-11-06/ai-startups-creating-new-gig-economy-for-bankers-lawyers]. Deccan centralizes operations in India to tightly manage quality—an approach founder Reddy says is more effective than spreading operations across a hundred countries.

India’s influence in the global AI value chain is increasingly evident. The country acts as a crucial supplier of talent and training data, although the development of core frontier models remains concentrated among U.S. and a handful of Chinese companies. Still, Deccan is starting to source select talent from additional markets—including the U.S.—for specialized needs such as geospatial data and semiconductor design.

From its inception, Deccan set out to be a “born GenAI” company, focusing on high-skill tasks, in contrast to older data-labeling firms that began with computer vision. Over the last year, Deccan has grown tenfold, now posting double-digit millions in annual revenue. Approximately 80% of its revenue is concentrated among its top five clients, underscoring the focused nature of the frontier AI sector.

Tags: Deccan AI, inteligjenca artificiale, trajnimi i modeleve AI, India, financim startup, vlerësimi i të dhënave