PROJECTS
03
Context Data · Applied research lab
The context layer for machines that act.
We research and build the infrastructure AI systems need to work safely on real data: governed access, rigorous evaluation and full traceability. Released as open source.
/projects
Instruments for AI in production.
CD-01 · DATA ACCESS● LIVE
Release candidateApache 2.0Open →
InterLock
Governed access to production data for AI agents: role and policy, redaction, approval for risky writes, and a full audit record.
CD-02 · EVALUATION○ BUILDING
In development
Cleanroom
Ties a movement in an evaluation score to the dataset or pipeline change that caused it, with contamination checks and lineage.
CD-03 · OBSERVABILITY○ BUILDING
In development
Chronicle
Distributed tracing for agent runs: every step, tool call and decision, so a failed run can be reconstructed and understood.
/research
Three open questions, one per layer.
ACCESS
What may an agent read and change, down to the row and column?
EVALUATION
When a score moves, which data change caused it, and was the test clean?
OBSERVABILITY
When a run fails, can every step and decision be reconstructed?
/lab
How we work.
01 / MEASURED
Every capability carries a status and its evidence.
02 / OPEN
Infrastructure in the data path should be inspectable.
03 / PROVEN
Deployed on real systems before we call it done.
SUPPORTED BY
