Mathematician · From Babylonian scribes to Fields medalists and AI-assisted proof: the profession that turns hard questions into permanent certainty, one theorem at a time.
Darker cells mean a higher score for this topic on that metric.
LessMore
Last reviewed Sources & creditsMedia creditsMethodology
Research mathematicians spend their days trying to prove statements no one has proved: reading papers, computing examples, filling blackboards, and writing up the small fraction of ideas that survive. Most also teach and supervise students. Applied and industrial mathematicians build and analyze models — of markets, machines, diseases or data — where the mathematics must answer to messy reality. Very little of either job is arithmetic.
For research and university positions, yes, effectively everywhere: the PhD is the credential that proves you can produce new mathematics. Outside academia the picture is looser — trading firms, tech companies and government agencies hire strong bachelor's and master's graduates — and unlike medicine or law, nothing legally stops anyone from proving theorems without any degree. Ramanujan never completed one.
Sector matters far more than seniority. US mathematicians earned a median of about $116,000 in 2023 (BLS); a full professor in Germany typically earns roughly €80,000–110,000; a junior CNRS researcher in France starts near €30,000. The outliers are in finance, where senior quantitative researchers at top trading firms can clear several times a professor's salary — a small, selective slice of the profession.
AI systems reached gold-medal level on International Mathematical Olympiad problems in 2025, and proof assistants like Lean can now verify research-level arguments. What machines still cannot do is decide which questions matter, invent the definitions that make a problem tractable, or carry responsibility for mathematical meaning. Most researchers expect the job to be transformed — closer to working with a powerful collaborator — rather than eliminated.
Alfred Nobel simply left mathematics out of his 1895 will; the popular story of a feud with the mathematician Gösta Mittag-Leffler has no documentary support. The field built its own honors instead: the Fields Medal, awarded every four years since 1936 to mathematicians under forty, and the Abel Prize, awarded annually by Norway since 2003 with a purse comparable to a Nobel's.
Seven problems the Clay Mathematics Institute named in 2000, each carrying a $1 million prize: the Riemann hypothesis, P versus NP, the Navier–Stokes and Yang–Mills problems, the Birch and Swinnerton-Dyer and Hodge conjectures, and the Poincaré conjecture. Only the Poincaré has been solved — by Grigori Perelman, who refused both the prize money in 2010 and the Fields Medal before it.
The core skill is not calculation — machines took that over decades ago — but the construction of airtight arguments about precisely defined objects, and the harder meta-skill of choosing which arguments are worth attempting. Working mathematicians describe their days honestly as being stuck: Andrew Wiles compared research to fumbling through dark rooms for months until you find the light switch, then moving to the next dark room.
The temperament the job selects for is therefore unusual: comfort with sustained failure, since almost every approach tried on a hard problem dies; obsessive precision, since one flawed line invalidates a hundred pages; and enough taste to abandon a doomed direction early. Speed, the quality schools reward, matters least — several of the field's greatest figures described themselves as slow.
Building complete, gap-free chains of logic from definitions to conclusion — the profession's product, held to a standard of certainty no other field uses.
Judging which questions are both important and attackable with existing tools — the skill that most separates careers, and the one advisors say is hardest to teach.
Returning to the same wall for months or years without progress; Wiles spent seven years on Fermat, and the daily texture of research is failed attempts.
Turning private understanding into papers referees can verify and colleagues can build on — the field's whole memory is written proof, and badly written results get lost.
Running experiments in SageMath, Mathematica or Python to generate data and test conjectures, and increasingly writing formal proofs in Lean — the modern lab bench.
Most research posts are professorships: lecturing, supervising doctoral students and passing the craft down the apprenticeship chain is half the actual job.
The day opens with the overnight arXiv listings — new preprints in one's area posted worldwide — plus referee requests and co-author email across time zones.
The protected morning block on the current problem: filling pages and blackboards, computing examples, attacking the same lemma that failed yesterday. Hardy held that four creative hours a day was a mathematician's limit.
Departmental lunch and tea are working institutions, not breaks — half of collaboration starts as a corridor remark, and strong departments defend the ritual fiercely.
Lecturing undergraduates, then meetings with doctoral students — checking a claimed proof line by line, or handing a stuck student a smaller version of their problem.
The research seminar — an hour of someone's new work, then questions — followed by chalk sessions with co-authors, where most joint papers actually get made.
Evenings go to writing and editing papers. The rest is genuine work too: Poincaré documented how solutions surface after conscious effort stops, and mathematicians deliberately load a problem before sleep.
Craft knowledge practitioners actually pass on — not motivation.
George Pólya's How to Solve It (1945), the best-selling manual of mathematical heuristics, codified the field's working method: if you cannot solve a problem, find a related easier one — a special case, an analogue, the problem with one condition dropped — and solve that first. Research mathematicians do this instinctively; Pólya made the instinct teachable, and the book has sold over a million copies.
Gauss discovered the prime number theorem's content as a teenager by hand-counting primes in tables, thousands per sitting, decades before anyone could prove it. The craft survives intact: modern researchers run computer experiments to build data, guess the pattern, and only then attempt proof. The maxim is that a good conjecture, honestly earned from examples, is half the theorem.
Alexander Grothendieck described two ways to open a nut: hammer and chisel, or immersing it in water for weeks until it opens by itself. His method — build the general theory so patiently that the hard problem eventually falls open with no visible struggle — rebuilt algebraic geometry and remains the deliberate strategy behind much modern mathematics: enlarge the framework instead of forcing the problem.
Henri Poincaré described working fruitlessly on Fuchsian functions for weeks, then having the solution arrive fully formed as he stepped onto an omnibus at Coutances, thinking of nothing mathematical. His prescription — saturate consciously, then let incubation work during rest — is deliberate technique across the field, which is why mathematicians defend walks, showers and sleep as working hours.
G. H. Hardy worked from breakfast to lunch and spent afternoons at cricket, holding that four hours of creative work a day is about a mathematician's limit and pushing past it produces errors to be undone tomorrow. Deep concentration on proof is physiologically expensive; the discipline is protecting the peak hours, not extending them.
Terence Tao's widely read career advice codifies the modern working method: strip a hard problem to its simplest nontrivial case, add every convenient assumption, and solve that — then remove the assumptions one at a time. He also counsels writing down partial progress and failed approaches, since knowing why the obvious attack fails is itself transferable capital.
Still the field's defining instrument: chalk's speed matches the pace of explained thought, and erasing invites risk-taking. The revered Japanese Hagoromo chalk survived its maker's 2015 closure only because a Korean firm bought the formula.
The universal typesetting system for mathematics, built on Donald Knuth's TeX (1978) — written because Knuth found his own book's galleys ugly — and Leslie Lamport's macros. Every journal, preprint and thesis in the field is written in it.
Software that verifies every logical step of a formalized proof. Lean's mathlib library, built by an open community since 2017, passed a million lines of formal mathematics and checked research-level theorems, including Peter Scholze's in 2022.
Mathematica, Maple, MATLAB, Magma and the open-source SageMath handle the symbolic and numerical experiments — factoring, plotting, matrix computation — that generate the examples and data from which conjectures grow.
The preprint server founded by Paul Ginsparg at Los Alamos in 1991, where essentially all new mathematics now appears before journal publication. Checking the morning's listings is the profession's shared daily ritual, and posting there timestamps priority.
Spending years on the Riemann hypothesis or Collatz conjecture with no strategy for extractable partial results is a recognized early-career killer; Paul Erdős said of Collatz that mathematics was "not yet ripe" for it. Wiles could gamble seven years on Fermat only because partial progress on modularity was publishable on its own.
Careers built on being the fastest in the room often crack at the research frontier, where everyone was the fastest in their room and problems take months, not minutes. The adjustment from speed to stamina defeats a meaningful share of doctoral students, and advisors watch for it more than for any gap in knowledge.
The field's absolute standard of correctness tempts researchers into polishing results indefinitely while priority slips away — the desk drawer of 90-percent-finished papers is a standing joke with a real body count behind it. The arXiv sharpened the race: an unposted result can be independently found and timestamped by someone else tomorrow.
Closest neighbours on the six-score profile — not the same field only.
The scholar paid to profess a subject in public — twenty-five centuries of teaching and discovery, from Plato's grove to the tenure track and the AI tutor.
AI-resistant 76 🎬Turns a screenplay into a finished film by deciding every shot, performance and cut, then persuades a producer, a studio and an audience the vision was worth the budget.
AI-resistant 73 🧩The scientist of mind and behavior, from Wundt's 1879 Leipzig laboratory to the therapy room — trained to hear what a person cannot yet say themselves.
AI-resistant 82 💱The scholar of scarcity — from Adam Smith's pin factory to the central-bank decision room, still asked to predict what no model fully captures.
AI-resistant 62 🔭The scientist who measures the universe, from Babylonian clay tablets to space telescopes, still deciding which flicker in the data is a discovery.
AI-resistant 70 ⚖️The person societies trust to decide: from Hammurabi's stele to AI-scored bail hearings, the hardest disputes still end in front of one accountable human.
AI-resistant 90Derives and tests the mathematical laws governing matter, energy, space and time, from a lone chalkboard to a 3,000-author particle-collider paper.
AI-resistant 65 🧬The scientist who studies life itself, from Linnaeus naming species by hand to editing genomes with CRISPR, still testing every idea against a living organism.
AI-resistant 64 🧪The scientist who makes and measures matter itself — from Tapputi's Babylonian perfume still to today's robot laboratories, still the one who decides what the spectrum means.
AI-resistant 70 📊Finds patterns and builds predictive models from data — a 2008 job title built on three centuries of counting, testing and visualizing evidence.
AI-resistant 38 🔭The scientist who measures the universe, from Babylonian clay tablets to space telescopes, still deciding which flicker in the data is a discovery.
AI-resistant 70 🧿Builds, measures and controls devices that exploit quantum states for computing, sensing, communication and materials research.
AI-resistant 78 🧫Uses clinical, trial and health-system data to generate reliable evidence for safer care, research and operational decisions.
AI-resistant 68