Deccan AI, a provider of high accuracy post-training data and RL environments for AI models, has secured a $25 million all equity Series A round led by A91 Partners, with participation from Susquehanna International Group and returning investor Prosus Ventures.
Deccan AI has closed a $25 million all equity Series A round, marking its largest and most visible funding to date. The round was led by A91 Partners with participation from Susquehanna International Group and returning investor Prosus Ventures. This brings the company’s total disclosed funding to $25 million across two rounds, following an earlier early stage round in May 2025.
What is Deccan AI’s main focus?
The capital will primarily fuel expansion of Deccan AI’s expert workforce, enhancement of its post training data pipelines, reinforcement learning (RL) environments, automated evaluation platforms, and EnterpriseOS agentic infrastructure. Founded by Rukesh Reddy and headquartered in Mountain View, California, with core operations anchored in Hyderabad, the company positions itself as a specialized provider of “super accurate” AI infrastructure. Its offerings target the post training phase of model development, where frontier labs and enterprises need reliable data to move beyond raw capability toward production grade safety, alignment, and real world performance.
Deccan AI’s core products address well documented gaps in current AI systems. It supplies multi modal datasets (audio, video, visual, document intelligence), coding repositories for training and evaluation, agentic trajectories (tool use, browser use, computer use), model alignment data (instruction-following, contextual awareness, trust and safety), and domain specific datasets in STEM, finance, and consulting. Complementing these are containerized RL environments with built-in verifiers, trace level checks, and end state validation crafted by subject matter experts. Its Helix platform delivers continuous monitoring, drift detection, and hybrid AI-human evaluation workflows, while EnterpriseOS embeds controllable agents directly into enterprise systems with human oversight. This full stack focus on accuracy, rather than scale alone, differentiates it in a market increasingly skeptical of hype driven benchmarks.
The timing of the round aligns with surging enterprise demand for post training services. As large language models proliferate, the bottleneck has shifted from pre training compute to high quality human feedback, RLHF pipelines, and rigorous evaluation that catch failures invisible in controlled demos. Deccan AI leverages a proprietary human plus AI vetting process and a deep pool of India-based domain experts (top 1 percent specialists) to deliver datasets and environments that frontier labs trust for alignment and safety work. This India-centric model provides both cost efficiency and access to specialized talent in a region rapidly emerging as an AI infrastructure hub, enabling the company to scale quality without the overhead seen in pure U.S. or Europe based annotation platforms.
Investor composition signals strong conviction in this thesis. A91 Partners, known for backing high growth Indian and global tech plays, leads the round, underscoring confidence in Deccan AI’s ability to dominate the high margin segment of AI data and evaluation infrastructure. Susquehanna International Group brings quantitative rigor and long term capital, while Prosus Ventures’ repeat participation reflects validation of early traction with enterprise and frontier model customers. The round’s structure, all equity, no secondary components, indicates founders and early backers are prioritizing growth over liquidity, consistent with the capital intensive nature of building defensible data moats and RL tooling.
Recommended: BusRight Raises $30 Million In Funding
Strategically, the funding accelerates Deccan AI’s competitive positioning against peers in the post training and agent evaluation space. It directly rivals human expert platforms like Mercor by emphasizing verifiable accuracy metrics and domain expert curation over generic crowd sourced labeling. With 125 employees and active research presence at conferences such as NeurIPS, the company is investing in proprietary studies that expose model weaknesses, feeding back into better datasets and environments. This closed loop approach creates a flywheel: more accurate data improves models, which in turn demand even more sophisticated evaluation and RL tooling.
The $25 million infusion positions Deccan AI to capture a larger share of the expanding AI infrastructure market. Enterprise adoption of agentic systems and physical intelligence applications will require precisely the kind of multi step, verifiable trajectories and safety aligned data Deccan AI specializes in. The company’s hybrid human-AI evaluation layer and on-premise EnterpriseOS offering also hedge against regulatory scrutiny around model reliability and data provenance. Potential risks remain, automation could eventually erode parts of the data labeling value chain, and competition from in-house teams at hyperscalers may intensify, but Deccan AI’s focus on irreplaceable expert judgment and production grade verifiers creates a defensible niche.
This Series A validates the market’s recognition that super intelligence without super accuracy is commercially unviable. Deccan AI’s latest raise equips it to scale the unglamorous but essential backend infrastructure that turns experimental models into trustworthy enterprise systems, solidifying its role in the next phase of AI deployment.
Please email us your feedback and news tips at hello(at)www.dailycompanynews.com
