LH2 AI Labs provides data and training infrastructure for AI labs, combining licensed institutional datasets, expert talent, and reinforcement learning environments. The company focuses on information and workflows that are typically absent from public web data, partnering with organizations to prepare proprietary sources for AI training. Its data spans software engineering, company operations, medical records, audio, and finance and legal materials, including code repositories, clinical imaging and notes, business documents, speech, calls, filings, and contracts.

LH2’s dataset pipeline licenses data at its source, removes identifying information, standardizes formats, preserves provenance, and uses domain experts to verify the results. It delivers machine-readable, training-ready corpora. The company also builds simulated environments that mirror real software and workflows, where AI agents can perform tasks and be evaluated. Examples include checkout and payment systems, CRM tools, hospital EHRs, and IDE and CI pipelines. Across the process, expert contributors can structure and verify data, write solutions, rank model outputs, and grade agent performance.