EXPOSED
Roughly half the job is screen work AI now does well or better: literature synthesis, sequence and structure analysis, statistical modeling, figure generation, manuscript and grant drafting — and AlphaFold-class models erased much of the structure-prediction workload outright. What holds is the bench: designing and troubleshooting assays, purifying proteins, running cell culture and instrumentation, and deciding which result is real versus artifact. There is no license or signature requirement, so nothing regulatory protects the analytical tier, and lab automation plus AI compresses how many bodies a project needs.
Headcount grew steadily across the period.
Median pay $94,490 → $127,410 +7.9% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
+5.8% 35,600 → 37,600 on the projections basis
Exposed, but growing
AI can already do a lot of these tasks, and the BLS still expects +5.8% more of these jobs by 2034. Demand for the output is growing faster than the work is being automated away — the mechanism BLS gives for software developers, and the combination people most often misread as an error.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~2,900 openings a year on average, including replacing people who leave.
ChemistScientistBiochemistBiophysicistToxicologistBiological ChemistProtein BiochemistChemistry ScientistClinical BiochemistClinical ResearcherMolecular BiologistPhysical BiochemistResearch BiochemistBiophysics ResearcherForensic ToxicologistFormulation ScientistBiochemistry ScientistPharmaceutical ScientistAnalytical Research ChemistClinical Laboratory ScientistProtein Biochemistry ScientistPostdoctoral Scholar (Postdoc Scholar)Postdoctoral Associate (Postdoc Associate)Postdoctoral Researcher (Postdoc Researcher)
Holding it up: judgment & accountability . Weakest point: liability shield .
Mixed — a routine tier and a judgment tier At 10 the split is real: the dry-lab half — homology searches, docking runs, kinetics curve-fitting in Prism, RNA-seq pipelines, methods sections — is already model-assisted, while the wet half (cloning a construct that won't ligate, chasing a Western that keeps ghosting, tuning an HPLC gradient for a protein that aggregates at pH 7) still needs a person at the hood, which is why it sits mid-band rather than at 5 or at 15.
Some physical or field component 12 reflects hands that matter but in a controlled room: BSL-2 hoods, -80 freezers, ultracentrifuges, mass spec sample prep, plate readers and cryostats are physical and dexterity-dependent, but it's all bench-top in a climate-controlled facility with fixed protocols, not the uncontrolled field conditions that push a score past 13.
No licence, no signature requirement 3 is nearly floor because no state licenses a biochemist — no PE stamp, no board certification, no signature on a regulatory filing; a PhD and a PI's approval get you to the bench, and GLP or IRB requirements attach to the institution and the study director, not to your personal credential.
Meaningful discretion 13 sits at the top of the discretion band because you decide whether an outlier is biology or a pipetting error, whether an off-target effect kills a candidate compound, and which mechanism the data actually supports — high-consequence calls, but ones that go through PI review, lab meeting, and peer review before anything irreversible happens, unlike a clinician's bedside decision.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (10/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (3/20) is whether the law requires a licensed human to sign. Trust premium (6/20) is whether buyers specifically pay for a person. Judgment and accountability (13/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 22 of this occupation's 44 points (50%).
Embodiment (12/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
Nuclear Engineers EXPOSED
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 58/100, still EXPOSED.
As AI absorbs literature synthesis, structure prediction and figure/manuscript drafting, the residual role concentrates on assay design, artifact adjudication and troubleshooting one-off wet-lab failures — genuinely two-tiered work where the judgment tier is the bench. Watch for job postings shifting from 'computational biochemist' to 'assay development / protein science' titles.
If journals and funders extend authorship-accountability rules (ICMJE-style, NIH's 2023 ban on AI in peer review) into a requirement that a named scientist vouch for data provenance and AI-assisted analysis in every submission, and institutions enforce it through research-integrity offices after fabricated-image scandals, the role formally owns the call on what is real.
For the subset working under FDA GLP/21 CFR Part 11 or GMP, an FDA or EMA requirement that a named qualified scientist personally attest to AI-derived analytical data and model-generated CMC/potency assay results would create a signature point. The FDA's January 2025 draft guidance on AI in regulatory decision-making for drugs, and EMA's AI reflection paper, are the live documents; a credentialed-attestation clause would be the specific thing to watch.
If the field's center of gravity shifts further toward hard-to-automate modalities — cryo-EM sample prep, membrane protein purification, primary cell and organoid culture, single-molecule biophysics rigs — where instrument-specific hands-on troubleshooting dominates, embodiment weight rises. Contrast with cloud-lab/Emerald-style automated pipetting, which cuts the other way.
Weak route only: CRO and biotech clients paying for a named PhD to sign off on assay validity because a wrong go/no-go call costs a clinical program — but this is priced as expertise, not human-ness, and AI-assisted competitors underprice it.
The limit. No licensure exists for the title, so liability protection can only arrive through the pharmaceutical regulatory perimeter and reaches only the industry subset — academic and discovery-stage biochemists get nothing from it. Headcount compression from automation is the dominant force regardless of dimension scores: the remaining judgment tier can be real and still require far fewer people.
| Boston-Cambridge-Newton, MA-NH | 13,370 | $131,550 +3% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 3,050 | $134,090 +5% |
| New York-Newark-Jersey City, NY-NJ | 2,200 | $109,620 -14% |
| San Francisco-Oakland-Fremont, CA | 1,380 | $165,050 +30% |
| Worcester, MA | 1,060 | $125,420 -2% |
| San Diego-Chula Vista-Carlsbad, CA | 710 | $121,740 -4% |
| Los Angeles-Long Beach-Anaheim, CA | 580 | $131,700 +3% |
| Trenton-Princeton, NJ | 520 | $124,300 -2% |
| Jacksonville, FL | 40 | $174,880 +37% |
| San Francisco-Oakland-Fremont, CA | 1,380 | $165,050 +30% |
| Knoxville, TN | 30 | $149,140 +17% |
We have no reported case of a named organisation automating this occupation. Not one deployment, not one announcement.
That is worth saying out loud next to a score of 44. The verdict above is about what the work exposes — what current AI could do to these tasks. It is not a claim that anyone has done it. For this occupation those two things have come apart completely: the capability argument is on this page, and the evidence column is empty.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.