Expert data for AI agents
Verified coding and business-workflow data for AI agents.
Senior engineers and practicing professionals write the data. Our platform proves it: tests run in a sandbox, graders are attacked before they ship, and a person approves every answer.
Proof, not promises
- 50
- RL environment tasks, each verified
- 12
- coding tasks proven in our sandbox
- 4
- grader attacks every task must defeat
- 82
- sample records you can download today
What we deliver
Launch focus: coding and AI-agent data, and practice environments for business software. The same platform runs every domain.
Verified coding tasks
- Experts:
- Senior engineers write a problem, starting code, a reference solution and hidden tests.
- You get:
- Tasks our sandbox proved: the tests fail before the fix and pass after, twice.
Business-software environments
- Experts:
- Finance and operations professionals design realistic tasks in practice copies of real apps.
- You get:
- RL environments with automatic scoring, each task attacked before it ships.
Preference and grading data
- Experts:
- Compare two AI answers, or score one against a rubric, and explain why.
- You get:
- Signals that teach a model what “better” means, including when to stop and ask.
Expert-written answers
- Experts:
- Practitioners write the answer they would want from an expert assistant.
- You get:
- Gold-standard examples to train on, reviewed by a senior expert.
How we check quality
No answer reaches you until it passes automatic and human checks. Rejected work goes back to the expert.
Tests that prove themselves
Every coding task runs in an isolated sandbox: fail on the starting code, pass twice on the reference.
Graders we try to break
Each environment task must defeat four blanket strategies before it ships.
Known-answer tasks
Gold tasks are mixed in unannounced; misses lower an expert's quality score.
Second opinions
Several experts per task where it matters, with agreement measured and reported.
A person approves every answer
Senior reviewers approve, return or reject; nobody reviews their own work.
No AI-written answers
Speed, known-answer and writing-pattern signals, and removal on the first confirmed case.
Security and independence
- Your data stays inside the platform: no shared documents, no public links.
- Experts see only the task in front of them, never your company name.
- Submitted code runs in our own locked-down containers, not a third-party sandbox.
- Canonset is independent of every AI lab.
For experts
Paid, flexible project work that uses what you know. Apply once, see the pay before you start, get paid weekly. No fees, ever.