Black Mesa ResearchThe unfair advantage, quietly deployed.
We build private AI systems around how your business actually works. No cloud. No leaking your data. Just measurable wins, compounding every week.
NO SALES TEAM
SMALL BY DESIGN
PRIVATE
We build private AI systems around how your business actually works. No cloud. No leaking your data. Just measurable wins, compounding every week.
Work
The workflows, the workarounds, the parts nobody documented. Strategy first — we study where the hours go before we build anything.
Every week has the same bottlenecks. We automate the repetitive ones with AI built around your operation — and the wins show up on the record, week after week.
Tech
Strategy first: we study where the hours go, then build on open-weight models — the right one for each job — on our hardware or yours.
Private inference
On hardware we control — or yours. Not on a public cloud. Not in anyone's training pipeline.
Open-weight, kept current
Llama, Qwen, Mistral — we pick per job. When a better version ships, we move to it. No lock-in.
A written data charter
No third-party training. No third-party egress. Your data stays yours.
A human holds the last word
The model drafts. A person approves. On the record.
Introducing Resonance
Resonance is our private networking layer. It puts your people, your sites, and your systems on the same secure wire — across offices, shops, and the field — without a public cloud in the middle. It is how you reach the private inference pipeline, and it is built on the same rule as everything we make: private by architecture, not by policy.
Nobody in the middle
Traffic flows device to device over encrypted tunnels. Not through a vendor's network. Not through ours, when it doesn't have to be.
End to end, always
Every hop is encrypted on the wire. When a relay is needed, it passes ciphertext it cannot read.
Only people you approve
Devices join by invitation, one at a time, under your control. Leave anytime — your network, your rules.
Resonance is in private beta — access is issued to our customers.
A managed model service on top of Resonance: curated open-weight models — text, code, vision, reasoning — running on our hardware, metered per user, reachable only through the mesh. Works with the tools you already use.
Tuned for security operations: log analysis, anomaly detection, alert triage. Built from open-weight foundations, hardened for the job, and running entirely inside your perimeter. The system watching your network should not be the biggest leak in it.
Contact
No mail app? Copy our address: [email protected]
FAQ
Depends on the job. After one conversation we'll scope a pilot and give you a number. You'll know the cost before any work starts.
The environment you approve. Private inference on our hardware or yours, under a written data charter: no third-party training, no third-party egress.
A small team can only answer for so much work. Fewer clients, better work — that's the whole trick.