Deccan AI wins $25M for GenAI post-training vs Mercor

Deccan AI, a San Francisco–based AI post-training startup, secured a $25 million Series A to challenge Mercor in GenAI model refinement. The all-equity round was led by A91 Partners, with Susquehanna International Group and Prosus Ventures also participating. Deccan AI focuses on the “born GenAI” post-training phase rather than basic data labeling. Its services include capability enhancement for coding and reasoning, tool/API integration training, expert feedback and evaluation via its Helix suite, and RLHF reinforcement-learning environment work. Founder Rukesh Reddy says quality remains “unsolved,” and error tolerance is “close to zero.” Commercial traction is early but measurable: Deccan AI currently serves about 10 customers (including Google DeepMind and Snowflake), runs several dozen active projects, and targets a double-digit million-dollar annual revenue run rate. The company also emphasizes an India-centric execution model: roughly 125 employees in Hyderabad plus a network of 1M+ contributors in India (5,000–10,000 monthly active), aiming to improve quality control through concentrated talent. Traders’ takeaway: while Deccan AI is not a crypto-native project, AI infrastructure spend can affect broader risk sentiment and tech-sector narratives. Still, the direct link to crypto market stability is limited, so the near-term impact on prices is likely modest. Crypto context keywords: GenAI post-training, AI evaluation, RLHF, and Deccan AI’s $25M funding round.