Deccan AI Raises $25 Million to Scale High-End Post-Training AI Services From India Hub

New updates have been reported about Deccan AI.

Deccan AI has secured a $25 million all-equity Series A round to expand its specialized post-training data and evaluation services for advanced AI models, positioning the company as a key infrastructure provider to frontier labs and enterprises. The round, led by A91 Partners with participation from Susquehanna International Group and Prosus Ventures, will fund growth of its India-centered expert workforce and product suite as demand accelerates for reliable, production-grade AI systems.

Founded in October 2024 and headquartered in the San Francisco Bay Area with major operations in Hyderabad, Deccan AI focuses on higher-skill post-training work, including reinforcement learning environments, expert feedback, and evaluations for coding, agents, and tool-using models that integrate via APIs. The company serves frontier labs and enterprise clients through offerings such as its Helix evaluation platform and an operations automation system, and it reports a tenfold revenue increase over the past year, reaching a double-digit million-dollar run rate.

Deccan AI currently has about 10 customers and manages a couple of dozen active projects at any time, with reported clients including Google DeepMind and Snowflake, and roughly 80% of revenue concentrated in its top five accounts, underscoring its dependence on a small number of large AI buyers. The company employs around 125 staff and coordinates a network of more than 1 million contributors, of which 5,000 to 10,000 are active monthly; roughly 10% hold advanced degrees, and pay ranges from $10 to $700 per hour, with top contributors earning up to $7,000 per month.

Management emphasizes that quality and speed in post-training are now critical constraints for AI labs, as tolerance for errors in evaluation and reinforcement learning is effectively near zero and timelines for high-quality data delivery are often measured in days, not weeks. To maintain control over quality, Deccan AI has chosen to concentrate most contributors in India rather than spread operations across dozens of markets, while selectively adding niche expertise from the U.S. in areas such as geospatial data and semiconductor design.

This India-centric strategy reinforces the country’s role as a core supplier of AI talent and training data rather than frontier model development, giving Deccan AI access to deep technical skills at scale while keeping operational oversight tight. At the same time, the company faces competitive and reputational pressures in a crowded segment that includes Scale AI, Surge AI, Turing, and Mercor, all vying for high-value labeling, evaluation, and reinforcement learning workloads.

Founder Rukesh Reddy positions Deccan AI as a “born GenAI” firm that avoided legacy low-skill labeling work and instead built its model around complex, domain-specific tasks tailored to large language models and emerging world-model systems in robotics and vision. For executives, the funding round validates Deccan AI’s positioning as a specialized post-training partner to top-tier AI developers, but it also highlights customer concentration risk, execution demands on quality and turnaround time, and the need to sustain differentiation as the AI infrastructure market matures.