AfterQuery

AfterQuery is an applied research lab that works with domain experts to capture how professionals reason through real-world tasks, then structures that knowledge into training data for frontier foundation models. The work addresses a gap between models that produce answers and the decisions, tradeoffs, and context that characterize actual professional work. The lab's research focuses on identifying where models break down in specific professional contexts and why those failure modes occur.

Its data products include supervised fine-tuning datasets with prompt-response pairs and chain-of-thought traces, rubric and verifier-based reinforcement learning, tool-calling RL environments built on APIs and MCP servers, computer-use and browser-use environments with expert-demonstrated trajectories, RLHF comparison pairs, code generation data, deep research tasks, multimodal data, and custom evaluations and training datasets. Loss analyses examine model failure modes to inform dataset and environment design. Off-the-shelf datasets are available across capability areas, and the data catalog can be accessed by request.

The products are aimed at AI research labs and companies developing foundation models and agents. The lab works across professional domains including medicine, law, finance, and engineering.