Deccan AI
Category: AI Infrastructure
A comprehensive AI infrastructure platform providing pristine, human-curated data for frontier AI model training, evaluation, and post-training services including RLHF and reinforcement learning environments. Deccan AI was founded in 2024. The company is led by Rukesh Reddy. Based in Mountain View, United States. Team size: 200+. Total funding raised: $25.0M. Latest round: Series A ($25.0M, Mar 2026). Key investors include A91 Partners, Prosus Ventures, Susquehanna International Group (SIG).
AMW Analysis
Deccan AI, founded in 2024 and based in Mountain View, operates as an AI infrastructure platform focused on human-curated data for frontier model training, evaluation, and post-training work, including RLHF and reinforcement learning environments. Its stated value proposition rests on a large expert-contributor base, including a substantial share of PhD-qualified specialists, delivered through its proprietary Databench platform with multi-layered quality assurance, and covering data generation, evaluation, and RL environments under one offering.
The company's news flow centers on a $25 million Series A led by A91 Partners, with Susquehanna International Group and Prosus Ventures also participating, reported across multiple items between March and April 2026. Coverage tied to the round describes Deccan as serving customers including Google DeepMind and Snowflake, with reporting citing a 2025 revenue run rate and prior business growth. The round is described as intended to expand existing platform tools, including Helix and EnterpriseOS, toward higher-accuracy model evaluation and post-training services. A separate late-2025 item describes the launch of "Deccan AI Experts," a network aimed at expanding its contributor pool to support reinforcement learning and enterprise AI data needs.
AMW analysis, generated from 3 tracked news signals.
- Founded
- 2024
- Headquarters
- Mountain View, United States
- Team size
- 200+
- Total funding
- $25.0M
Value proposition
Delivers production-ready AI models through a unique combination of 1M+ expert contributors (including 53,000+ PhD-qualified specialists), proprietary Databench platform with multi-layered quality assurance, and comprehensive coverage across post-training needs (data generation, evaluation, RL environments)
Products and solutions
Databench Platform - Proprietary AI data curation and allocation system, STARK RL Environments - Code-based containerized reinforcement learning environments with five-step verifiers, Helix Enterprise Eval Suite - Hybrid human-plus-automated evaluation system with continuous agent monitoring, EnterpriseOS - Operations automation platform for enterprise AI deployment, Multi-Modal Training Data - Custom datasets across text, image, audio, video, and document intelligence, Coding Solutions - Private code repositories for training and evaluation (SWE-Bench compatible), Text-to-SQL - Natural language to SQL conversion for database access, Domain-Specific Solutions - STEM, finance, consulting, and specialized domain data, Agentic Solutions - Tool use, browser use, computer use, and multi-step execution trajectories
Unique value
Operates the world's largest curated expert network of 1M+ contributors including 53,000+ PhD-qualified specialists, combined with proprietary Databench platform that allocates work based on expertise, performance history, compliance, and language depth. The platform embeds multi-layered quality architecture including peer review, gold-label systems, simulation-based trials, and anomaly detection - achieving 'top 1% raters' quality standard.
Target customer
Frontier AI labs (Google DeepMind), enterprise technology companies (Snowflake), AI research institutions, and organizations deploying production AI systems requiring high-accuracy training data and evaluation
Industries served
Artificial Intelligence & Machine Learning, Enterprise Software & Technology, Financial Services & FinTech, Software Development & DevOps, EdTech & Education, Healthcare & Life Sciences, Research & Development, Data Science & Analytics
Technology advantage
Three-pronged competitive advantage: (1) Proprietary Databench platform with Human + AI Quality Playbook ensures near-zero error tolerance in production AI systems; (2) Strategic India-based talent pool provides cost-effective access to elite technical expertise (11,000-12,000 deployed specialists scaling to 50,000); (3) End-to-end post-training infrastructure spanning data generation, evaluation systems, and reinforcement learning environments. The platform supports the full AI lifecycle from code generation and agentic workflows to physical intelligence for robotics, with purpose-built GenAI features including linters, code compilers, LaTeX editors, and LLM-based validators.
How they differentiate
Leverages India's elite technical talent pool with curated network of 1M+ contributors (53,000+ PhD-qualified experts) through proprietary Databench platform, multi-layered quality assurance with peer review and gold-label systems, and end-to-end post-training infrastructure spanning data generation, evaluation, and reinforcement learning environments
Main competitors
Scale AI (Meta-owned), Mercor, Surge AI
Key partnerships
Google DeepMind - Frontier AI lab customer for model training and evaluation, Snowflake - Enterprise technology customer for AI model development, Prosus Ventures - Strategic investor (seed and Series A participation), A91 Partners - Lead investor for $25M Series A round, Susquehanna International Group (SIG) - Series A investor, Deccan Experts Network - 1M+ contributor ecosystem including 53,000+ PhD-qualified specialists, Frontier AI Research Community - Active participants and presenters at conferences like NeurIPS
Notable customers
Google DeepMind, Snowflake, Majority of Magnificent 7 tech companies, Fortune 500 enterprises
Major milestones
Founded in October 2024 by Rukesh Reddy, Seed funding from Prosus Ventures in May 2025, Built network of 1M+ contributors with 53,000+ PhD-qualified experts, Achieved double-digit million-dollar revenue run rate, Series A funding led by A91 Partners in March 2026
Growth metrics
10x business growth in first year, double-digit million-dollar annual revenue run rate, 80% of revenue from top 5 customers
Market positioning
Premium AI infrastructure provider bridging frontier AI labs and enterprise deployment, focusing on high-accuracy, production-ready AI systems with near-zero error tolerance
Geographic focus
United States (San Francisco Bay Area headquarters), India (Hyderabad major operations center)
Patents and IP
No registered patents publicly disclosed as of latest update. The company's competitive moat is built on proprietary processes, curated expert networks, and specialized platform capabilities rather than patent protection.
About Rukesh Reddy
15+ years across Citi, Monitor Group (now Monitor Deloitte), JP Morgan, Soul AI, and 360 ONE Wealth. B.Tech from IIT Bombay (2002-2006) and MBA from IIM Ahmedabad (2007-2009).
Latest news about Deccan AI
- Deccan AI secured $25 million in a Series A led by A91 Partners, with SIG and Prosus Ventures joining. The capital will turbo‑charge its enterprise‑grade AI platform, expanding tools like Helix and En
- Deccan AI, a post-training data startup serving Google DeepMind and Snowflake, raised $25M Series A led by A91 Partners with SIG and Prosus Ventures. The round signals that human-in-the-loop quality r
- Prosus-backed Deccan AI has launched "Deccan AI Experts," a network for advanced AI model training and evaluation, aiming to expand its skilled annotator pool to 50,000 contributors. This initiative a
More AI Infrastructure companies
Official website: https://www.deccan.ai
