Drug Discovery Scientist at Mercor

Drug Discovery Scientist | $60–90/hr | Worldwide Remote

Join a cutting-edge AI evaluation initiative where your deep scientific expertise shapes how frontier AI models reason through complex drug research and development challenges. You'll design rigorous, realistic research problems that require genuine scientific reasoning — not pattern matching — to solve.

About the Role

This role supports an evaluation framework for frontier AI models in drug R&D. You will craft realistic research scenarios embedded in carefully constructed evidence worlds, where the model must work through incomplete, indirect, and occasionally misleading data to reach a defensible scientific conclusion. The answer is always present in the evidence — but never stated outright.

What You'll Deliver

• Challenge Design: Identify the hidden conclusion and construct a forcing chain that makes one answer scientifically inevitable.
• Data Room Construction: Assemble a source-grounded data room where every decisive value traces to a real, verifiable reference.
• Grading Document: Write a grader that evaluates submitted memos on scientific substance — not style or method.
• Validation: Run verification and calibration on each task before submission.
• Revision: Own the rework process if a reviewer requests changes.

You are the final authority on all scientific and difficulty decisions within your tasks.

Who We're Looking For

• Working expertise in mechanistic enzymology, structure-guided peptide or antibody design, fragment-based discovery, or structure-based medicinal chemistry.
• PhD or equivalent industry track record in a relevant discipline.
• Ability to pose problems whose answers you can defend rigorously, yet which strong solvers cannot shortcut.
• Comfortable evaluating reasoning quality, not just final answers.

Commitment & Tooling

• Minimum 20 hours/week, up to 40 hours/week.
• Work is conducted in a browser-based studio with Claude Code.
• A Claude Max subscription is required and fully reimbursed.