Frontier Physics Researcher: Golden Reference Solutions

Mercor; Paris; À distance; EUR 95 000 - 161 000; Temps partiel

Résumé du poste

Mercor is seeking expert physics researchers to author and verify golden reference solutions for frontier physics benchmarks. You will solve problems end-to-end, audit others’ solutions, and adjudicate between attempts to produce human-verified data for evaluating models.

Ideal candidates have PhD or postdoc experience, strong English writing, and hands-on LaTeX, Python, Jupyter, and SymPy skills. Commitment is about 10 hours per week over an 8–10 week window, with asynchronous collaboration.

Qualifications

• PhD or postdoc in a relevant subfield with publications.
• Hands-on familiarity with multiple canonical methods in the subfield.
• Strong written English and ability to document reasoning clearly.
• Experience creating verifiable, testable solutions with references.

Responsabilités

• Solve research-level physics challenges end-to-end with derivations and code.
• Decompose tasks into sub-problems requiring physical reasoning.
• Create Python templates with auto-grading for symbolic/numeric answers.
• Audit solutions for correctness, scope, and method soundness.
• Adjudicate between parallel solver attempts and select golden references.
• Document reasoning, tolerances, and verification test cases.

Connaissances

LaTeX

Python

Jupyter

SymPy

English writing

Audit adjudication

Formation

PhD in relevant subfield

Postdoc experience

Senior PhD student

Outils

LaTeX

Python

Jupyter

SymPy

Description du poste

Mercor is seeking expert physics researchers to author and verify golden reference solutions for frontier physics benchmarks. You will solve problems end-to-end, audit others’ solutions, and adjudicate between attempts to produce human-verified data for evaluating models.

Ideal candidates have PhD or postdoc experience, strong English writing, and hands-on LaTeX, Python, Jupyter, and SymPy skills. Commitment is about 10 hours per week over an 8–10 week window, with asynchronous collaboration.