San Francisco: Deccan AI has raised $25 million in a Series A round to scale its AI post-training and evaluation services for frontier labs and enterprises. The all-equity round was led by A91 Partners, with participation from Susquehanna International Group and Prosus Ventures.
Founded in October 2024, Deccan sits in a category that rarely makes headlines but powers the entire AI industry. While labs like OpenAI and Anthropic build core models in-house, the post-training work data generation, evaluation, and reinforcement learning is increasingly outsourced as companies push to make systems reliable in real-world use.
Deccan handles that layer. Its services range from improving model coding and agent capabilities to training systems to interact with external APIs. It also builds reinforcement learning environments and runs evaluations for frontier lab customers including Google DeepMind and Snowflake.
The company is headquartered in San Francisco with a large operations team in Hyderabad. It employs around 125 people and draws on a network of over one million contributors including students, domain experts, and PhDs. Between 5,000 and 10,000 contributors are active in any given month.
The scale of that contributor network matters. Post-training is not a compute problem, it is a human judgment problem. Getting AI models to behave reliably in the real world requires expert feedback at volume. That is what Deccan’s India-based workforce provides, and it is why the model is attracting serious venture backing less than 18 months after founding.
Deccan also offers two enterprise products: Helix, an evaluation suite, and an operations automation platform. It currently serves around ten customers and runs several dozen active projects simultaneously.
The work is also expanding beyond language. As models move into so-called world models, systems that better understand physical environments. Deccan is positioning itself to support training for robotics and vision systems as well.
The $25M will fund headcount growth, product development, and expansion of its contributor network as demand for post-training services accelerates alongside the broader AI infrastructure boom.
Previous
