Example · PhD, Physical Sciences and Engineering
73% of the work a physical sciences PhD student is heading toward still needs a person.
AI now writes analysis code and first drafts. The instrument work and your name on the result stay with you.
Make my planA typical week at the middle pace. Your plan uses your own tasks.
Is a physics or engineering PhD worth it with AI?
In a typical week at the middle pace, 53% of this work needs your judgment and 20% is work you do using AI. AI could do about 27%. Funding, visas and the faculty market will shape your next three years more than any model. AI changes something narrower.
Read more
The code, the first-draft analysis and the literature review in your week are getting cheap, while the bench, the instrument and the judgment about which measurement to run next still need you. In a lab-heavy program, about half of a week is work that holds up.
So the plan does two things. It turns the hands-on work you already do into evidence an employer can see. And it makes you the person who checks the model's output, which is a skill labs and companies are short on right now. Neither move asks you to leave your field or become an AI researcher. If your PhD is fully computational, read the last section first, because your numbers are different.
The week, task by task
Who does each hour of a typical 40-hour week.
See every task
Needs your judgment.
- Hands-on experiments: synthesis, fabrication, cleanroom, sample prep, measurement18%
- Designing experiments and choosing the next measurement8%
- Building, maintaining and troubleshooting instruments8%
- Presenting and defending results: group meeting, committee, conferences, design reviews7%
- + 3 smaller tasks11%
AI does the slow parts. You still own it.
- Analyzing data, fitting models, making figures10%
- Setting up and running simulations: DFT, MD, finite element, HPC jobs8%
- Ordering, scheduling instrument and beam time, group logistics2%
AI could do most of it.
- Writing analysis code, fitting scripts and data pipelines11%
- First drafts of papers, proposals, reports and fellowship essays8%
- Literature searches, reading and summarizing papers5%
- Grading, writing problem-set solutions, answering routine course email4%
Why some tasks still need a person
Growing films, synthesizing samples, cleanroom and bench work
This takes presence. Someone has to be in the room with the sample. Self-driving labs exist, and they still need people to judge whether a result is real. When outside researchers re-examined an autonomous lab's 43 claimed new materials, they concluded "no new materials have been discovered in that work" (Leeman et al., PRX Energy, January 2024).
Keeping an instrument running, and knowing its habits
This takes context and presence. O*NET lists maintaining laboratory instruments and troubleshooting malfunctions as a core chemist task. An instrument's quirks are written down nowhere except the log of the person who keeps it alive.
Deciding what to measure next
This rests on context. Experiment design depends on what failed last month, what the tool can actually do and what your PI will fund. A model can list options, but it does not know your lab.
Lab safety and the chemical hygiene plan
Your lab hands you this responsibility. OSHA's lab standard requires "Designation of personnel responsible for implementation of the Chemical Hygiene Plan" and training on it (29 CFR 1910.1450).
Standing behind a result in front of a committee, a referee or a design review
You hold this accountability. Your name is on the paper and the thesis, and you answer the question "how do you know?"
Good to know
What is hard about this field, and what you can do about it.
The hard truth
Academic jobs pay about half what industry pays.
Their academic starting pay is the lowest of any field NSF reports. Staying in academia is a real financial choice.
Sources (1)
- In 2024, new physical science PhDs expected a median $65,140 in academia and $65,000 in postdocs, versus $120,000 in industry. NSF NCSES, Survey of Earned Doctorates, 2026-02
What careers come after a physical sciences PhD?
Plan A is the best move in the field you are studying for.
Focus your studies
The main focus. Most weekly sessions build toward it.
- Each week
- 4 sessions
- Ready in
- One semester for the record and applications
Why it fits
Why you. If you run a deposition, synthesis or characterization tool now, those skills transfer straight into semiconductor process, integration and metrology, battery and energy materials, or aerospace materials and testing.
The gap. You don't yet have an industry internship or a written record of instrument ownership.
The trade-off. Industry hiring runs in cycles. Some aerospace and defense roles require "U.S. person" status under ITAR, the US arms export rules, which covers citizens, lawful permanent residents and certain protected individuals (22 CFR 120.62), or a security clearance, which is generally limited to US citizens (Executive Order 12968, section 3.1(b)). Pay varies by sector and region.
2 other moves that fit
- Do 3 to 12 months of thesis research at a national labReady in three weeks for the application
- Keep a coding job open as a backup: data science, scientific software or quant researchReady in about 13 weeks
What's a plan B for physics PhDs?
Plan B is meant to be different from this job. It was picked because skills from this role carry over.
Get ready for a new kind of job
A second option. A newer job AI is creating that these skills already fit.
Standing behind a result in front of a referee trains the exact rigor platforms want when grading AI physics and chemistry answers.
- Each week
- 1 session
- Outlook
- Hiring now
- Pay
- $50 to $110 an hour (contract)Three legal AI trainer contracts in 2026; the top rate requires a JD.
2 other options
- LLM evaluation data scientistYour fitting scripts and instrument data work are a natural start on the computational exit toward evaluating AI models.
- AI policy analystNational labs and fellowships need scientists who can explain AI's limits, and deciding what to measure next is that kind of judgment.
In your own plan you can switch to any of the 47 roles.
The 13 weeks
Small steps that add up. One plan for both: most sessions build Plan A, one a week builds Plan B.
By week 13 you will have
AFor Plan A
- An SCGSR or NDSEG application submitted
- A methods note comparing a model with your own measurements
- Ten internship applications sent
BAnd for Plan B
- A timed practice record
- An expert task pack for your portfolio and platform applications
- 1See where your time goes
- 2Check fellowship deadlines
- 3Submit your first application
- 4Submit SCGSR and start a log
You finish withA yes or no on owning an instrument
- 5Test a model against your data
- 6Log where models go wrong
- 7Write a committee decision memo
- 8Build your industry list
You finish withA yes, a no or a "not eligible"
- 9Finals: a lighter week
- 10Grading: a lighter week
- 11Break: draft a methods note
- 12Break: feedback and applications
- 13Share your work and apply
You finish withTen internship applications in total
Three courses from the plan
- Office of Science Graduate Student Research (SCGSR), 2026 solicitationDOE Office of Science · Stipend up to $3,600 a month; round-trip travel paid
Yes, Weeks 2 to 4.
- NDSEG Fellowship, FY2027Department of Defense · $3,600 a month ($43,200 a year) plus up to $5,000 travel
Only if you are a US citizen in your first two years of graduate study.
- NSF Graduate Research Fellowship (NSF 26-526)NSF · $37,000 stipend; $16,000 a year to the school
No. It is closed after your first year.
Checked 2026-10-13.
The fuller picture
Risks beyond AI: highFederal funding and access risk
Graduate positions, postdocs and national lab jobs are paid from federal science budgets. FY2026 held, and the FY2027 request asks for deep cuts again. For international students, visa fees and lab access rules add a second layer that has nothing to do with AI.
Where this could be wrong
- The blend assumes a lab-heavy student. A purely computational or theory student has far more of the week in code, simulation and analysis. Without bench work, less than a third of the week holds up.
- Self-driving labs may move faster than we scored, mainly in powder synthesis and solution chemistry. Thin-film, cleanroom and device work is slower to change. Capability is uncertain here. Lab budgets and institutional policy set the pace.
- The postdoc and sector shares behind the blend come from AIP's report on physics PhDs from the classes of 2015 and 2016, published in 2019. It is the latest full breakdown we found, and the mix may have shifted.
- Funding figures come from AAS, Astrobites and Science News summaries of agency and congressional documents. FY2027 is not final as of October 13, 2026.
Also matched: phd student physics, physics phd, chemistry phd. Research checked 2026-10-13. Page updated .