Deccan AI Raises $25M Series A for AI Data Evals
Deccan AI, a US-India-based provider of human-curated SFT/RLHF data, RL environments, evaluation suites, and enterprise agentic solutions, has raised $25M in Series A funding led by A91 Partners. The company uses a proprietary Human + AI vetting process with a 500k+ specialist talent pool to deliver pristine datasets and tools for accurate AI models. The capital will fuel scaling post-training data services, R&D, and enterprise infrastructure.
India Talent Powers Data Surge
The round closes amid rising investment in AI annotation platforms. Mercor raised $350M Series C at a $10B valuation in late 2025, focusing on global experts including India. Deccan AI's India-centric model with 1M+ vetted experts addresses quality demands as agentic AI proliferates. Scale AI has secured $1.6B total funding for similar data labeling. This funding validates Deccan's differentiation in super-accurate RLHF and evals.
Post-Training Errors Threaten Agents
Quality issues persist in AI post-training data, where error tolerance nears zero. Agents fail in production due to hallucinations and unreliable feedback loops from weak datasets. Current solutions lack rigorous verification for complex agentic workflows. Enterprises demand production-grade evals amid the shift beyond chatbots.
Human-AI Vetting Ensures Pristine Data
Deccan AI's proprietary platform combines real-time linters, gold-standard QA, and 100+ in-house quality associates. Products include expert-curated datasets for agentic, coding, and multimodal tasks; STARK RL Gym for enterprise-grade training; and Helix for continuous monitoring and drift detection. Unlike broad labeling platforms, Deccan emphasizes domain-specific accuracy via IIT/NIT alumni experts.
EnterpriseOS Targets Regulated Ops
EnterpriseOS delivers bespoke agentic workflows for compliance, fin-crimes, HR, and support. Built with embedded governance, it suits regulated environments. As Rukesh Reddy, Founder and CEO, noted:
"Quality remains an unsolved problem – tolerance for errors in post-training is close to zero."
This approach powers reliable models for frontier labs like Google and Snowflake.
Investors Signal AI Infra Bet
A91 Partners leads with India scale-up expertise, marking their first AI deal. Prosus Ventures, a prior seed investor, doubles down on AI agents. Susquehanna International Group (SIG) brings quant data prowess. The mix signals conviction in Deccan's US-India model for enterprise AI.
Kaushik Anand from A91 added:
"Deccan is building the essential infrastructure for the next decade of software."
Annotation Market Hits $13B
The AI annotation market stands at $2.51B in 2026, projected to reach $13.11B by 2033 at 31% CAGR per ResearchAndMarkets. Demand surges for RLHF/SFT amid agentic and multimodal model growth. Deccan's research benchmarks like Anthar and image-gen studies highlight gaps in current evals.
Rivals Scale Labeling Operations
Scale AI dominates with broad integrations but faces capacity pressures. Surge AI bootstraps RLHF focus, eyeing $1B round. Mercor's expert platform mirrors Deccan's India leverage but lacks integrated RL gyms. Deccan carves a niche in verification-heavy post-training tools.
Banking Veteran Leads Team
Founder Rukesh Reddy, ex-Head of Citigold at Citi managing billion-dollar revenue, brings IIT Bombay and IIM Ahmedabad credentials. The team features ex-Google, Amazon, AWS leaders for GTM and AI engineers from SigTuple and GE Healthcare. This blend scales operations from 220 employees with SOC2/HIPAA certifications.
Enterprise Expansion Accelerates
Post-raise, Deccan plans aggressive hiring across US and India roles. A new Bengaluru office targets Fortune 500 via GCCs. Revenue hits double-digit millions at 10x YoY growth, with majority Mag7 clients.