Engagements, described by what they had to solve

Building automation

Controller platform for motorised shading

The situation

A controller line ran firmware that only one former employee could build. There was no remote update path, and every configuration change needed an engineer on site with a laptop.

What we did

We rebuilt the controller software on embedded Linux with a reproducible image, added signed A/B updates with automatic rollback, and gave every unit a hardware‑backed identity. An installer application commissions units over Bluetooth with no internet on site, and a web console manages whole portfolios centrally.

Where it left them

Configuration and firmware changes now reach installed units without a site visit, and the build can be reproduced by any engineer on the team.

Industrial robotics

Fleet control for autonomous mobile robots

The situation

Vehicles from more than one supplier, each with a different control interface, and no single view of where the fleet was, what it was carrying or how much charge remained.

What we did

We built a control and telemetry layer over ROS 2 and CANopen that presents a mixed fleet through one interface, added battery and charging management with predictive scheduling, and delivered a dispatcher console showing live position and task state.

Where it left them

Mixed fleets are dispatched as one, and charging is scheduled around the work rather than interrupting it.

Commerce operations

Multi-marketplace seller platform

The situation

Orders, stock and returns spread across several marketplaces, reconciled by hand in spreadsheets. True margin per order was only visible weeks after the fact, if at all.

What we did

One operations platform that syncs listings and inventory across channels, ingests settlement reports and reconciles them automatically against orders, and reports per‑order margin including fees, returns and shipping as it happens.

Where it left them

Manual reconciliation was removed from the daily routine, and pricing decisions are now made against current margin rather than last month's estimate.

Manufacturing

Production tracking for a made-to-order plant

The situation

Job status lived on paper travellers. Supervisors could not tell which orders were late until a customer called, and the office chased the floor by phone.

What we did

We built production tracking around the way the floor already worked, with job stages captured at the machine and progress pushed to supervisors and customers through the messaging apps they already use rather than a portal nobody would log into.

Where it left them

Order status is visible as it changes, and chasing moved from phone calls to exceptions that surface on their own.

Consumer products

Device platform and companion app for a connected product line

The situation

A hardware product with growing sales, no fleet management, and a setup flow that failed often enough on real home networks to generate returns.

What we did

We added a device platform with provisioning, telemetry and staged update campaigns, then rebuilt the setup flow around the conditions that actually caused failures — dual‑band routers, captive portals and interrupted first connections.

Where it left them

Units can be updated after sale, and setup failures became a measured number with a path to reduce it rather than an unexplained support cost.

Infrastructure

Managed hosting and deployment for a software portfolio

The situation

A growing set of applications on hand‑configured servers, deployed manually, with no consistent monitoring and recovery steps that lived in one person's memory.

What we did

We standardised on containerised deployment with automated pipelines, put monitoring and alerting in front of every service, and documented and rehearsed the restore procedure until it was routine.

Where it left them

Deployments became repeatable and unremarkable, and recovery steps are written down and tested rather than recalled under pressure.