AI

AI training data startup AfterQuery nabs $30M investment

Artificial intelligence data provider AfterQuery Inc. has raised $30 million in funding at a $300 million valuation.

The startup disclosed in its Thursday announcement of the deal that Altos Ventures was the lead investor. Y Combinator, The Raine Group and BoxGroup chipped in as well.

San Francisco-based AfterQuery provides datasets that AI developers use to train their models. The 14-month-old company says its customer base includes “every leading AI lab.” Its annual recurring revenue topped $100 million shortly before the funding round.

AI training datasets often comprise prompt-response pairs, or records that consist of a sample prompt and a human-written answer. AfterQuery says that its datasets also include a step-by-step overview of the thought process behind each prompt response. That information makes it easier for AI models to apply the lessons they learn from a prompt-response pair to other tasks.

Frontier AI models usually go through not one but several rounds of training. AfterQuery customers can commission datasets optimized for specific training steps such as the reinforcement learning stage. In that phase of an AI project, the model being trained completes a set of sample tasks and receives feedback on how well it completed them. That feedback, which is often provided by a second AI model, helps improve the quality of prompts. 

AfterQuery generates training datasets with the help of nearly 100,000 developers, attorneys and other professionals. Those experts can provide not only natural language prompt responses but also other types of data. A company building a coding agent, for example, could commission files that demonstrate how developers go about debugging software bugs. AfterQuery also provides multimodal training data. 

The company sells its training datasets alongside other technical assets. It offers evaluation suites, software toolkits that measure the effectiveness of AI training sessions. It can also put together a development environment with MCP servers, application programming interfaces and other components. Customers can use such environments to hone their coding agents’ programming capabilities.

AfterQuery also provides other kinds of AI training sandboxes. It can put together a virtual environment that simulates an organization’s employee workstations and browser configuration. According to the software maker, such sandboxes can be used to develop AI agents that automate business tasks.

The startup will use the proceeds from its funding round to grow its expert network and workforce. The company also plans to enhance its enterprise business, which sells offerings such as custom AI agents.

AfterQuery is at least the third AI data startup to have raised funding in the past month. Deccan AI Inc., which provides datasets for the post-training phase of AI projects, raised $25 million in late March. AI training environment developer Deeptune Inc. closed a $43 million round a few days earlier.

Image: Unsplash