Design before code
Anything substantial starts as a written design, reviewed until it stops changing, then deliberately simplified. Deleting a paragraph costs less than deleting a subsystem.
Infrastructure systems engineering · Timișoara, RO
The layer everything else runs on.
We build infrastructure software — hypervisors, clustered storage appliances, and the networking and trust that hold them together when something fails.
Where we work
The full depth of the stack.
Infrastructure problems rarely respect layer boundaries. Following one all the way down — and fixing it at the layer that actually owns it — is the job.
Surface
-
Operator surface
The interfaces and documentation people actually run the system with. If an operator can’t tell what it is doing, it isn’t finished.
-
Services & APIs
Clear contracts between services, and orchestration that unwinds cleanly when a step fails halfway through.
-
Clustering & availability
Systems that stay up when a machine leaves badly, and agree with themselves again once it comes back.
-
Virtualization
Hypervisor platforms end to end, from the life of a virtual machine to the performance of what runs inside it.
-
Storage & data integrity
Capacity, snapshots and replication built so that a copy is provably a copy — up to mirrors that lose nothing at all.
-
Platform, network & trust
The ground the rest stands on: the operating system, the network it sits on, and the certificates that keep it all trusted.
Hardware
Runs through every layer
Observability
Systems that can be understood while they are still running, instrumented before anyone needs it — and dashboards built for whoever is on call at 3 a.m. rather than for the demo.
AI-assisted engineering
Models given safe, scoped access to live systems, and used daily for design review, triage and implementation — with a human accountable for everything that ships.
Practice
How the work actually goes.
Validated on real systems
Designs are proven against real hardware, and the findings go back into the document before implementation starts. A whiteboard has never caught a firmware quirk.
Written for the operator
Interface contracts, playbooks and documentation ship with the feature rather than as a follow-up ticket. The people who run the system are the ones we design for.
Owning the failure
Production incidents get a root-cause file, a mitigation, customer comms in plain language, and a recovery procedure support can run without us on the call.
Contact
Working on something down here?
Design reviews, hard problems in virtualization and storage, or long-term ownership of a platform.
dan@glidercode.ro