What the work involves
Each task starts with a bioactivity record — a compound, an assay readout, and a protein target mapped to a UniProt ID. Your job is to go back to the source: pull the paper or patent, read the methods closely enough to know what protein construct was actually assayed, and decide whether the assigned accession is right. That means paying attention to species, isoform, subunit composition, complexes versus single chains, orthologs, chimeric or tagged constructs, and cases where the assay measured a downstream functional response rather than direct binding to the named protein.
When an assignment is wrong, you classify the error against a structured taxonomy, propose the correct accession, and write a short evidence-grounded note — usually a few sentences citing the specific figure, table, or methods line that justifies your call. Consistency matters more than volume. These labels feed AI drug discovery models, so a confidently wrong correction is more costly than a flagged ambiguity.
What the platform screens for
Mercor's screen is AI-led and looks for bench reality, not résumé keywords. Expect probing on assay formats you have personally run (SPR, TR-FRET, radioligand binding, kinase or GPCR functional assays), how you interpret IC50 versus Ki versus KD, and whether you can articulate the difference between an agonist/antagonist framing and an activator/inhibitor framing. Follow-up questions get more specific, not less, so answer with concrete detail: constructs, buffers, controls, what went wrong and how you diagnosed it. Fluency with UniProt and adjacent workflows — target validation, selectivity profiling, SAR support, HTS triage — is checked directly.
Logistics
- 10–20 hrs/week, fully remote and largely async; you choose your hours
- 1–2 month minimum engagement, extension described as likely
- Immediate start; U.S.-based candidates only
- Requires being currently bench-active in a research, scientist, or associate scientist role at a biotech, pharma, or CRO
- Pay observed at $50–70/hr; final rate depends on credentials and platform calibration and is not guaranteed