As the race to build more capable and reliable AI systems accelerates, a new layer of infrastructure is emerging — one focused not on training models, but refining them for real-world use. For Deccan AI, this shift represents a significant opportunity.
The startup has raised $25 million in Series A funding, positioning itself as a key player in the rapidly growing market for AI post-training services. The round was led by A91 Partners, with participation from Susquehanna International Group and Prosus Ventures.
Founded in 2024, Deccan AI is building infrastructure that supports the final stages of AI development, where models are evaluated, refined, and prepared for deployment.
The Rising Importance of Post-Training in AI Development
While leading AI labs such as OpenAI and Anthropic continue to develop large-scale foundation models, much of the post-training work is increasingly being outsourced.
This phase includes:
- generating high-quality training and evaluation data
- running performance testing and validation
- building reinforcement learning environments
- improving model capabilities such as coding and agent behavior
As AI systems move into production environments, the tolerance for error becomes extremely low — making post-training a critical step in ensuring reliability.
According to Rukesh Reddy, quality remains one of the biggest challenges in this stage of development, with even small errors having a direct impact on model performance.
Building Infrastructure for AI at Scale
Deccan AI operates at this post-training layer, working with frontier labs and enterprise customers to refine AI systems.
Its services range from improving model reasoning and coding capabilities to enabling interaction with external systems such as APIs — a key requirement for enterprise use cases.
The company also offers proprietary tools, including an evaluation suite and an operations platform designed to automate parts of the AI development workflow.
Its client base includes major technology organizations such as Google DeepMind and Snowflake, reflecting growing demand from both AI labs and enterprise customers.
A Workforce Model Built Around Expert Talent
A defining aspect of Deccan AI’s approach is its reliance on a large, distributed network of contributors.
The company employs around 125 full-time staff while tapping into a pool of over one million contributors — including students, domain specialists, and PhD-level experts.
At any given time, thousands of contributors are actively working on projects, providing the specialized input required for high-quality post-training tasks.
Unlike traditional data labeling firms, Deccan AI focuses on higher-skill work, requiring deep domain expertise and precision.
India’s Growing Role in the AI Value Chain
While Deccan AI is headquartered in the U.S., much of its operational workforce is based in India, particularly in Hyderabad.
This strategy reflects a broader trend: India is increasingly becoming a global hub for AI training and evaluation talent.
Competitors such as Turing and Mercor also source talent from the region, though often across multiple countries.
Deccan AI, however, has taken a more concentrated approach — focusing heavily on India to maintain quality control and streamline operations.
At the same time, the company is expanding into other markets, including the U.S., for niche expertise in areas such as semiconductor design and geospatial data.
Competing in a Rapidly Expanding Market
The market for AI training and post-training services has grown alongside the rise of large language models and generative AI systems.
Established players such as Scale AI and Surge AI compete alongside newer startups, including Turing and Mercor.
What differentiates Deccan AI is its focus on post-training complexity, where precision, speed, and domain expertise are critical.
The company has reported rapid growth, scaling significantly over the past year and reaching a multi-million-dollar revenue run rate, driven largely by a concentrated group of enterprise and AI lab customers.
Balancing Speed, Scale, and Quality
One of the biggest challenges in post-training is balancing speed with accuracy.
AI labs often require large volumes of high-quality data within tight timelines, creating pressure to deliver quickly without compromising quality.
At the same time, the sector has faced scrutiny over working conditions and compensation models, given its reliance on distributed contributors.
Deccan AI has positioned itself differently by focusing on higher-value tasks and offering a wider earning range for contributors based on expertise.
A Strategic Bet on the Future of AI Infrastructure
As AI systems evolve beyond text into more advanced capabilities — including robotics, vision systems, and “world models” — the need for high-quality post-training will only increase.
Deccan AI is positioning itself at the center of this shift, building infrastructure that enables AI systems to move from experimental models to reliable, production-ready tools.
The company’s latest funding round underscores investor confidence in this emerging layer of the AI stack — one that is becoming essential as enterprises and AI labs alike push toward real-world deployment.