Diagnostic Radiologist - AI Trainer

Mercor; Houston (TX); On-site; USD 90,000 - 140,000; Full time

Job summary

Mercor is seeking practicing radiologists to define what “correct” looks like in AI-evaluated radiology reads across CT, MRI, and ultrasound studies. You will craft reference reports and apply established measurement conventions such as BI-RADS, Lung-RADS, and RECIST.

Applicants should be board-certified or board-eligible radiologists currently reading studies, with experience across multiple subspecialties and willing to review AI-drafted reads for accuracy and safety.

Qualifications

• Board-certified radiologists or board-eligible candidates with active practice.
• Experience reading de-identified imaging studies across modalities (CT/MRI/US/PET-CT).
• Familiarity with structured reporting systems (BI-RADS, Lung-RADS, RECIST, CAD-RADS, ASPECTS).
• Ability to review AI-generated reads and identify inaccuracies or unsafe conclusions.

Responsibilities

• Read de-identified studies and write reference reports or specific findings.
• Apply standard measurement conventions and structured systems in practice.
• Review AI-drafted reads and flag inaccuracies, unsupported conclusions, or safety concerns.
• Contribute challenging cases to improve model performance.

Skills

Reading radiology studies

Structured reporting

AI model evaluation

Critical thinking

Education

Board-certified radiologist

Board-eligible radiologist

Job description

We're working with a leading AI research lab to build rigorous evaluations of how well AI reads real radiology studies, and we need practicing radiologists to define what "correct" actually looks like. AI can describe an image fluently; whether it gets the clinically load-bearing call right — the measurement by the correct convention, the interval change against a prior, the structured category in a borderline case — is where expert judgment is scarce and where you come in.

This is not image labeling. It's the reasoning you already do every day, captured as a reference standard that trains and tests frontier models.

What you'd do

• - Read de-identified studies (CT, MR, radiography, ultrasound, PET/CT, mammography, and more) and write the reference report or the specific finding.
• - Apply the structured systems and measurement conventions you use in practice (RECIST, Lung-RADS, BI-RADS, CAD-RADS, ASPECTS, and others)
• - Review AI-drafted reads and mark what's inaccurate, unsupported, or unsafe.
• - Contribute the kinds of cases that stump a strong model.

Who we're looking for

• - Board-certified or board-eligible radiologists (ABR, FRCR, EDiR, or international equivalent). Residents and fellows in later training may be considered.
• - Any diagnostic sub-specialty: neuro, chest, abdominal, MSK, breast, cardiac, nuclear medicine, paediatric, and others.
• - Currently reading studies, or recently in active practice.