Skip to content

Portfolio

Projects

Architecture and method demonstrated through technical scenarios — clearly identified as such.

About this section

We do not publish client cases without written authorization — nor "anonymized" versions that, in practice, identify the client by description. Until authorization exists, this section presents illustrative technical scenarios, always labeled as such, whose purpose is to show architecture and decision criteria, not to suggest delivered work.

Illustrative scenarios

Each scenario starts from a real operational requirement in the sector and walks through the reasoning to the architecture — including what was rejected and what that rejection cost. None represents a client, a contract or delivered work.

Illustrative case

Redundant link for a remote site without fiber

Requirement
Continuous supervision of a site 40 km from the operations center, with no fiber on the route and none planned.
What constrains the answer
The operation tolerates no outage longer than the failover time, and the region has heavy concentrated rainfall — which rules out any redundancy that depends on the same physical medium.

Architecture

  1. 01Primary path over licensed point-to-point radio, sized against the region’s rainfall percentile rather than its average.
  2. 02Satellite contingency, with automatic failover triggered by route loss — not by interface loss, which takes longer to notice.
  3. 03Independent monitoring of both paths: a link that went down without telling anyone is worse than a link that does not exist.
What was rejected, and what it cost (3)
Two radios of the same type on different frequencies
It is cheaper and looks like redundancy, but both degrade together in the same storm. Redundancy that shares a failure mode is cost without coverage.
Unlicensed-band radio
It removes licensing cost and lead time. In exchange, there is no recourse when interference shows up — and on a continuous supervision link, it shows up.
4G/5G as the contingency
It would be the fastest to deploy. It was ruled out here because of coverage at the site: a contingency that depends on signal nobody measured on site is an assumption, not a design.

Illustrative case

Segmentation between corporate and automation networks

Requirement
An industrial plant whose SCADA is reachable from the administrative network, with no automation asset inventory and vendor remote access already in place.
What constrains the answer
Production cannot be stopped to discover what is on the network, and firewall rules applied over an incomplete inventory take down a process — usually the most critical one, which is exactly the one nobody documented.

Architecture

  1. 01PASSIVE asset and protocol discovery first, via traffic mirroring. Nothing is injected into the automation network during the survey.
  2. 02Zone and conduit design based on what was observed, not on what the old diagram claims.
  3. 03Industrial firewall at the boundary, in observation mode before blocking — the log of what would have been denied becomes the list of what the inventory is still missing.
  4. 04Vendor access through a broker, with recorded sessions and approved windows, replacing the standing access that was already in place.
What was rejected, and what it cost (2)
Active scanning to build the inventory
It is faster and far more complete. It is also the classic way to knock over an old controller that does not expect a packet outside its protocol. The time saved does not pay for the risk of a stoppage.
Block first, open up as complaints arrive
It reaches the same final design in fewer steps. But it moves inventory discovery into production, and the first stoppage destroys trust in the whole project.

Illustrative case

CCTV sized by imaging objective

Requirement
Read plates at the vehicle entrance and recognize people at the loading dock, with footage that works as evidence rather than just as a record.
What constrains the answer
The previous design specified cameras by count and by area covered, without declaring what each point needed to resolve — a criterion that produces wide coverage and useless evidence.

Architecture

  1. 01Each point gets a declared objective: detect, recognize or identify. Sensor, lens, lighting and frame rate follow from it — in that order.
  2. 02Lighting treated as part of the imaging design, not as an installation item: a good camera in a dark scene delivers high-resolution noise.
  3. 03Retention calculated from the objective, not from the disk that happens to be available.
What was rejected, and what it cost (2)
Keeping the original camera count
The final design uses FEWER cameras. Cutting points is always the hardest conversation in the project, because it looks like scope reduction — and it is the opposite: the points that were removed resolved no declared objective.
High resolution at every point
It simplifies specification and procurement. It costs bandwidth, storage and licensing at points where motion detection was enough — and it does nothing for plate reading, which depends on lens and lighting, not on megapixels.

Questions about this document?

Get in touch