/ 01
Deterministic over probabilistic
If a security signal can be explained by a rule, a model is overkill. We score by weighted signal decomposition before we ever consider a learned classifier.
About
A lab, not a SaaS.
BlackGrid Labs is the R&D arm of Axis Meridi Technologies, a division of Veydros Group. We work on security primitives that the rest of the suite needs — and that the rest of the industry treats as either solved or unsolvable.
Operating principle.
Security tooling has drifted toward black-box scoring — a probabilistic model assigns a number, a vendor takes a fee, and no one downstream can explain why a particular message was flagged or a particular login was rejected. That is not security. That is dependency.
BlackGrid works on the opposite premise. Every primitive we ship is deterministic: same inputs, same outputs, every time. Every decision is auditable: a person who reads the rule book can reproduce the score on paper. Every output is reproducible: held-out evaluations are public when they’re ready, not when they’re scheduled.
We invest in primitives that don’t fit the Suite or OrangePeel yet — security, authentication, deterministic detection — where the conventional approach is opaque, probabilistic, or hostage to a vendor. Work moves from the bench into product when it’s ready, and not before.
/ 02
Reproducible, then public
Held-out validation comes first. Public release of methods, datasets, and benchmark numbers follows only when the lab is willing to defend them. Output cadence: two public works per quarter.
/ 03
Primitive, not product
We build small, sharp pieces — an email-auth scorer, a tenant-isolation harness, a liveness step-up — that compose into the products downstream. The boundary is intentional.
What ships, and when.
Research that does not ship is folklore. BlackGrid measures itself in two ways:
- Product integration. A primitive earns its keep by landing in IronGrid (our threat-intel platform), MyMeridi (the identity layer), AxisVantage (client portal), or OrangePeel (the D.A.E. engine). If five years pass without a primitive shipping anywhere, the track is killed.
- Public output. Two works per quarter — an internal draft, a public proposal, or a benchmark dataset. We publish a result only when an outside reviewer could reproduce it from the description.
Engage the lab
Want to collaborate
or pilot a primitive?
We work with a small number of operators per quarter on directed research, pilot integrations, and tenant-isolation audits. If your problem is one of our open tracks, the conversation moves fast.