Systems we have
built and shipped.

The problem

Energy arrived as a single monthly invoice. Nobody could say what a shift, a line or a product actually cost to run, so no process change could be proven to have helped. Reconciliation happened in spreadsheets, weeks after the fact.

What we built

• An RS-485 meter network across every major load centre, polled by ADAM
• An attribution model mapping each meter to line, shift and product
• LARA dashboards with automatic shift, daily and monthly reporting
• Deviation alerts against a modelled consumption baseline

What it found

Significant idle-load draw over weekends and shutdowns that had never appeared in aggregate billing, and a formation block consuming well above its modelled baseline — both corrected within the first quarter of operation.

The problem

Utilities were treated as always-on background infrastructure. Nobody monitored compressor efficiency until a unit failed, and compressed-air leakage — the most expensive avoidable loss in most plants — was entirely unmeasured.

What we built

• Pressure, flow and temperature sensing across all four utility systems
• Specific-power computation per compressor, trended against commissioning data
• Night-time and shutdown flow analysis to quantify leakage
• Drift alerts raised against each asset's own efficiency baseline

What it found

Substantial leakage traced to specific headers and drop legs, and one compressor running consistently outside its efficient load band — a sequencing fix rather than a capital replacement.

The problem

The plant knew what it drew and roughly what it discharged, with no reconciliation between them. Statutory reporting was assembled manually from meter readings taken by hand, which made both accuracy and timeliness a recurring risk.

What we built

• Flow metering at intake, treatment inlet and outlet, and final discharge
• A continuously reconciled water balance across every stage
• Effluent treatment plant monitoring with parameter alarms
• Automated statutory reports generated to the required format

What it found

A persistent unaccounted volume that reconciliation localised to a specific distribution section — invisible while intake and discharge were only ever compared at plant level.

The problem

Plant-level generation totals looked healthy, which is exactly how underperformance hides. A handful of degraded strings and one drifting inverter can disappear inside an aggregate that still looks broadly correct for the season.

What we built

• Inverter and string-level acquisition across the array
• Performance ratio normalised against measured plane-of-array irradiance
• Ranked underperformance list, refreshed continuously
• Soiling and shading losses separated from genuine equipment faults

What it found

Specific strings persistently below their irradiance-normalised expectation, and a quantified soiling curve that turned panel cleaning from a fixed calendar routine into a decision with a payback attached.

The problem

Defects were logged in books after the fact, with no binding to the weld parameters in force when they occurred. Root cause was therefore a matter of experience and argument, and the same defect modes kept recurring because nothing proved what caused them.

What we built

• Point-of-creation defect capture at the mill, bound to pipe and batch identity
• Live parameter binding — current, voltage, travel speed, wire feed, gas flow
• Correlation analytics across thousands of welds per campaign
• Pareto and cost-of-quality ranking by scrap and rework value
• Classification models trained on the labelled defect history

What it found

A strong, repeatable relationship between wire-feed drift and porosity that had never been demonstrable before. Control limits were tightened around the evidence rather than around convention.

The problem

Material movement across assembly was manual and difficult to schedule. Off-the-shelf guided vehicles wanted fixed infrastructure and stopped dead whenever an aisle was obstructed — which, on a working shop floor, is constantly.

What we built

• SLAM mapping and AMCL localisation robust to a changing environment
• Nav2-based global and local planning with dynamic obstacle avoidance
• Automatic docking and charge scheduling
• Fleet traffic management, mission allocation and priority handling
• Telemetry reported into LARA alongside every other plant asset

What it changed

Robots that re-route around a blocked aisle instead of waiting for it to clear, and a fleet whose utilisation, battery health and mission history sit in the same system as the rest of the plant rather than in a separate vendor portal.