Skip to content
Products

Data Built for the Gap in Your Model

Our engineers work alongside yours, from the task design through to the data itself. We'll dig into the problem with you and come back with a pilot sample built to your spec.

Recent Work

We Measure It Ourselves

Coding RL environments

Refactoring and unit-test tasks as RL environments. A real repo snapshot in an isolated container, returning a graded reward rather than a pass mark.

  • Calibrated to a 20 to 80% pass band
  • Graders hardened against gaming attempts
  • Rubrics grounded in Fowler and Ousterhout

TypeScript, Python, Go and Rust. Local renames through inheritance restructuring.

Computer use SFT

Desktop sessions captured as canonical episodes. One episode projects into SFT, grounding, preference, repair and evaluator sets.

  • Cross-app planning with source-of-truth conflicts
  • Long-horizon editing with verified staged artifacts
  • Repair tasks scored against a hidden error list

Negative examples must be rejected by validators before a package ships.

MCP tool-use environments

Executable agent tasks, each with a restorable initial state and a verifier that scores the resulting environment rather than the model's prose.

  • pass@5 up from 33.3% to 50.0%
  • avg@1 up from 18.7% to 25.3%
  • Open 9B model, one GRPO + LoRA epoch, five runs

Measured on a fixed 30-task set, filesystem subset only. Database tasks were not in this run.

What We Build

The Areas We Go Deep In

Tool Use & Long-Horizon Agents

Agents that plan across tools and systems, recover when a run goes sideways, and deliver work that checks out.

Computer Use & GUI Agents

Desktop, browser and mobile sessions captured step by step, with the environment state to verify every action.

Coding

Executable engineering tasks from real codebases, graded by hidden tests the agent never sees.

3D, Games & World Models

3D assets and executable 3D tasks, game footage and source code, and industrial CAD for spatial and world-model training.

Multimodal Reasoning

Visual reasoning data weighted to where frontier vision models still break.

AI for Science

Expert STEM problems and scientific computing environments, backed by question banks from K-12 through professional exams.

Off-the-Shelf Data

Data We Already Hold, Ready to License

Our collections grow and change frequently. Request the catalog for up-to-date figures.

Books & Papers

E-books, textbooks, theses and academic papers, plus bibliographic records and classical texts.

Q&A & Community

General and developer Q&A, forums and social discussion, medical Q&A, reviews and subtitles.

News

Official, self-media and industry news across print and online publishing.

Encyclopedia & Web

Encyclopedias, wikis, how-to and lifestyle sites, long-form articles and cleaned web crawls.

Professional Domain

Clinical guidelines, court opinions, patents, company filings and research reports, and published government reports, plus cybersecurity research material.

Code

Repositories, pull requests and issues, front-end projects, and competitive programming solutions.

Images

Art and photography, science image-text, and fashion and e-commerce images with annotations.

Video

General video and clips, video-text pairs, film and drama, and commercials.

Speech & Audio

Southeast Asian ASR corpora and large-scale multilingual audio.