Industries

Semiconductor programmes run to a schedule set by the fab, not by the calendar. A new facility moves from cleanroom construction and base build through tool install and hook up, ultrapure water and bulk gas distribution, and commissioning, before a single wafer runs. Each phase draws a different mix of disciplines and each has a handover the next phase cannot start without.

Once the fab is producing, the work changes character rather than stopping. Process engineering runs across lithography, etch, diffusion, thin film and CMP. Equipment engineering keeps tools inside their qualification windows. Test and ATE development, yield analysis and failure analysis close the loop between what the line produces and what the process was designed to deliver.

Product and package engineering, VLSI design and embedded software sit alongside the fab, where silicon becomes a part somebody else builds into a system. The constant across all of it is that a delay in one discipline stops every discipline behind it, and the qualified people who can close that gap are not available at short notice in most labour markets.

A network is built outdoors, one site at a time, and progress is counted in sites on air rather than in features shipped. The work here is physical infrastructure and field operations, not BSS or OSS software. A rollout opens with RF planning: coverage and capacity modelling, frequency and neighbour planning, then drive test and post build optimisation to prove the design against what the radio actually does once the antenna is on the mast.

Transmission and microwave links carry the traffic back, and where the backhaul is glass rather than air, fibre and outside plant crews handle route survey, ducting, splicing and FTTx distribution down to the customer premises. Every site needs land before it needs equipment, so site acquisition, landlord negotiation and permitting run months ahead of the civil works, and the towers, masts and rooftop structures that follow need riggers certified to climb them. Power systems, rectifiers, batteries and standby generation keep a site alive when the grid does not.

Once traffic is flowing, NOC and field operations own fault response, preventive maintenance and the swap programmes that follow each technology generation. Holding the sequence together is its own discipline, because a permit that lands late strands a crew, a tower and a switching window that were all booked around it.

A data centre is a building whose only purpose is to keep power and cooling running without interruption, and every decision about how it is designed, built and operated follows from that. Design and construction begin with the electrical and mechanical topology: utility intake, transformers, switchgear, generators, UPS, and the chilled water or direct expansion plant that removes the heat those loads produce.

MEP coordination and BIM resolve the clashes between those systems before they reach site, because a routing conflict found in the field costs weeks. Commissioning then proves the facility in five levels, from factory acceptance and component checks through system level testing to integrated systems testing, where the whole plant is loaded with heat banks and failed deliberately to confirm it behaves the way the design says it will. Level 5 is the last point at which a fault is cheap to find.

After handover the building enters critical facility operations and the tolerance narrows again, because switching, maintenance and every method statement are now performed on live equipment carrying customer load. Electrical supervisors hold the permits, HSE governs work near arc flash and confined spaces, and technicians run the rounds, alarms and escalations that keep availability where the contract says it should be.

A plant is judged on whether it can make the same part to the same tolerance, at rate, for years. Production and process engineering own that question directly: cycle time, yield, scrap, changeover and the process window each operation has to hold. Industrial engineering sets line balance, layout, standard work and capacity, while maintenance and reliability keep the assets available through planned intervention rather than breakdown response.

Automation and controls run underneath all of it, from PLC and SCADA to robot cells and vision systems, and tooling and mould engineering decides how quickly a design change becomes a part in a box. New product introduction is where those disciplines meet at once, through design for manufacture, process validation, pilot builds and the ramp to volume, followed by the sustaining work that keeps a mature line profitable. Planning and materials hold the schedule against supply that rarely behaves.

Quality is audited rather than asserted, against IATF 16949 in automotive, ISO 13485 in medical devices and AS9100 in aerospace. Above the plant floor sits a digitalisation layer that is now part of the operation rather than a project: SAP PP, QM and PM, MES and SCADA integration, OT and IT convergence, and predictive maintenance built on data the machines were already producing.

Oil and gas divides into three streams that share engineering disciplines but not their constraints. Upstream is drilling, wellheads and offshore or onshore production facilities, where access is limited, bed space is finite and weather governs the schedule. Midstream is pipelines, compression, pumping and storage, measured in kilometres of line, hydraulic modelling and integrity management across ground that other people own. Downstream is refining and petrochemicals, where a plant runs continuously for years and the calendar is built around the next turnaround.

Across all three, projects are delivered through an EPC contractor and the client team that holds the interface, so engineering is written to a specification, checked against codes and then reconciled with what the fabricator and the site can actually build. Piping and pipeline engineering, process and mechanical engineering, and instrumentation and controls carry most of that scope, with HSE embedded rather than adjacent: hot work, confined space, hydrocarbon inventory and permit to work discipline decide whether a job proceeds at all.

A project ends in commissioning and start up. The busiest point in the life of an operating asset is a turnaround, where a plant is shut down, opened, inspected, repaired and restarted inside a window measured in days, and every day beyond it is lost production.

In a regulated plant the batch is only half the output, and the evidence that it was made correctly is the other half. GMP manufacturing environments are built around that principle, from classified cleanrooms and the airlocks and pressure cascades that separate them, through the utilities that feed them: HVAC, purified water and water for injection, clean steam and compressed gases. None of it is accepted on the strength of a datasheet.

Commissioning and qualification prove each system against a written protocol, and installation, operational and performance qualification are executed, deviated and approved in sequence before any product depends on them. Process and automation engineering then has to hold the validated state, because a change to a recipe, a controller or a filter grade is a change to the file, and the change control record travels with it. Quality assurance and compliance own batch review, deviation and CAPA management, and the inspection that arrives at short notice.

Medical device manufacturing adds its own weight, where sterile barrier systems, joint replacement implants and instrumentation carry design history files. Underneath all of it sits data integrity: computer system validation, audit trails and electronic records that must stay attributable and contemporaneous, because a plant that cannot prove what it did has not made a product it can release.

Almost no enterprise estate is greenfield. What exists is layered: an ERP that has been in production for a decade, a set of integrations nobody documented, a cloud footprint that grew faster than the controls around it, and a server fleet ageing out through staggered hardware replacement cycles. Work in this sector is therefore as much about the seams as about the components.

Enterprise platforms carry the transactional core, so changing one means understanding finance, supply chain and plant processes as well as the software underneath them. Cloud brings landing zone design, identity, network topology and migration waves, followed by the cost discipline that keeps consumption honest once the migration is declared finished. Application development runs alongside, from requirements and API design through build, automated testing and release. Infrastructure holds the compute, storage, virtualisation, network and endpoint layers that everything above assumes will simply be there.

Service management is what the business actually experiences: incident and problem handling, change control, request fulfilment, and the service levels each of those is measured against. Data engineering turns operational records into pipelines and models that reporting can defend. Cybersecurity runs through the whole estate, across identity and access, vulnerability management, monitoring and a response plan whose quality is only ever discovered on the day it is needed.

Engineering disciplines travel. A civil, structural, mechanical, electrical or process engineer who can read a drawing set, hold a specification and defend a calculation is useful in a fab, a substation, a treatment plant or a hospital, and the sequence a project follows barely changes between them. Concept and basic design fix the scope. Detailed design turns it into the drawings, models and datasheets a fabricator can price and build to.

Procurement engineering and expediting decide whether long lead items arrive before the crews booked around them do. Construction is where residual error in all of that becomes visible, so site engineering, setting out and supervision exist to close the distance between what was drawn and what the ground allows. Mechanical and electrical packages are usually the last trades in and the first ones blamed, which makes coordination with the main works contractor a full time role rather than a standing meeting.

Planning owns the critical path, the look ahead and the progress measurement the contract is paid against. Quality runs through inspection and test plans, witness and hold points, and non conformance reports nobody enjoys raising. After handover the asset passes to facilities and maintenance, where planned intervention, statutory inspection and an asset register that reflects reality decide how long the thing actually lasts.

The functions behind a technical organisation are not generic versions of the same jobs done elsewhere. Finance in a project business accounts for work in progress rather than months alone: cost to complete, revenue recognised across contracts that run for years, capital cases defended per asset instead of per department, and an audit trail that has to survive scrutiny long after a site has closed.

Human resources works to competency frameworks rather than job titles, because a certificate that has lapsed removes a person from a live site the same day, and training records, appraisal cycles and employee relations all sit against that fact. Procurement qualifies suppliers, runs tenders, negotiates terms and then administers contracts through delivery, where a specification that was loose at award becomes a claim at completion.

Supply chain plans demand, holds inventory against lead times that move without warning, and clears goods through customs, freight and delivery windows that rarely align with each other. Administration keeps document control, facilities, travel and statutory records in order, which is invisible until the day it fails. What these functions share is that they are judged by the technical operation they support, so the people who can read a project schedule as fluently as a ledger are the ones worth keeping.