EXPOSED
This catch-all bucket covers people who calibrate, align, and repair scientific, industrial, and specialty instruments — spectrometers, gauges, optical assemblies, sensors, timing and measurement equipment — using hand tools, test benches, and traceable standards. Almost none of that is text-on-a-screen work: the value is in disassembling a one-off instrument, finding drift or a failed component, and bringing it back inside tolerance, which today's robotics cannot touch. The soft spots are the paperwork layer — diagnostic lookup, calibration certificates, service reports, parts research — plus the slow trend toward sealed, module-swap instruments that turn skilled repair into replacement.
Tasks largely resist digitisation. Stripping a spectrometer's optical bench, finding a 0.2° mirror misalignment, and re-shimming it against a traceable standard is a sequence no model can execute and no arm can improvise on a one-off instrument — the 16 rather than 19 reflects that diagnostic decision trees, manual lookup, and parts sourcing are already moving to software, and sealed modules keep shaving off the repair end.
Hands-on in uncontrolled environments. You are inside the housing with jeweler's screwdrivers, torque specs, alignment lasers and a scope probe, often at the customer's plant floor or lab bench where the instrument sits — 18 not 20 only because much of the fine work happens on a controlled bench rather than up a tower or in a trench.
Certification preferred, not legally required. No state licence gates this work; ASQ CCT, NIST-traceable training or an employer's ISO 17025 accreditation is what buys you credibility, and when a calibration is later found out of tolerance the accredited lab and its quality manager absorb it, not your signature on the cert.
Some relationship component. Labs and plant engineers keep calling the tech who knows their particular 15-year-old mass spec and will pick up when it drifts mid-run, but the purchase order is written for turnaround time and traceability, so a competent replacement keeps the account.
Meaningful discretion. You decide whether drift is a failing detector or a contaminated sample path, whether an out-of-tolerance instrument's prior measurements must be recalled, and whether to repair or condemn a $200k asset — real calls with expensive consequences, held under 14 because tolerance limits, manufacturer specs, and calibration intervals are written down before you arrive.
Has AI actually changed your work?