Life Sciences (Drug Discovery) - AI Trainer

Job Description

About the role

\n

DataAnnotation is looking for experienced drug discovery scientists to evaluate how frontier AI models handle real discovery and preclinical work: SAR analysis, screening cascades, DMPK and PK/PD interpretation, biomarker strategy, candidate selection. You bring the judgment you have built moving programs from target to IND. We bring the model outputs that judgment is needed to grade.

\n

In this role, you will design challenging, realistic tasks drawn from your own practice, such as a lead optimization decision memo, a screening triage workbook, a tox package summary, or a target validation review, run them through frontier AI agents, and evaluate what comes back against a professional standard.

\n

You will work with realistic professional files, the kind a scientist in your field actually handles: experimental data, protocols, records, reports and correspondence. Some you will assemble yourself; others will be provided. In every case the goal is the same: a task a competent scientist in your field would complete correctly and a frontier model currently gets wrong.

\n

This is not a traditional lab or analysis role. You will be helping build better AI by putting your knowledge to work in a structured, flexible, fully remote environment. The work is long form and self directed, and clear written reasoning matters as much as technical depth.

\n

Responsibilities

\n

\n
• Design challenging, realistic drug discovery tasks drawn from your own day to day workflows: the scenario, a prompt phrased the way you would brief a trusted colleague, and the supporting files an agent would need (assay result tables, SAR spreadsheets, DMPK summaries, study reports, program review decks, correspondence), using files you author yourself or files that are provided to you.
\n
• Run those tasks through frontier AI agents and evaluate the deliverable they produce (the workbook, memo, or slide deck) against the standard you would hold a colleague to.
\n
• Compare two model outputs on identical prompts and files, decide which performed better, and document where each fell short.
\n
• Write detailed grading rubrics that specify what a correct deliverable must contain, such as the right compounds advanced, the right liabilities flagged and the right calculations, and explain in writing why a response passes or fails each one.
\n
• Flag concrete failures with evidence: misread assay data, SAR conclusions the data do not support, PK parameters misinterpreted, ignored or fabricated files, unit and scaling errors, missed safety liabilities, and off brief interpretation of the ask.
\n
• Contribute across drug discovery and target biology, DMPK and preclinical safety, translational and biomarker science, biologics and advanced modalities, and CMC and process development, and review and refine tasks built by other experts.
\n

\n

Domain qualifications

\n

\n
• 3+ years hands on in a discovery or preclinical setting (pharma, biotech, or an academic drug discovery unit).
\n
• Depth of experience in at least one of: medicinal chemistry and SAR or lead optimization; computational chemistry and CADD; assay development and screening; DMPK, PK/PD, or ADME; preclinical safety and toxicology; translational and biomarker science; protein or antibody engineering; cell and gene therapy; CMC, formulation, or analytical development.
\n
• Working understanding of several of the other areas above, enough to know what those workflows involve and how they are run (for example, a medicinal chemist who also understands how a screening cascade is built and how DMPK data feeds a lead optimization decision), so you can assess work in adjacent areas and point out what was done correctly or incorrectly.
\n
• Familiarity with drug discovery workflows end to end: how a program moves from target identification and validation, through hit finding and hit to lead, into lead optimization and candidate selection, and on into IND enabling studies.
\n
• Nice to have: experience across more than one modality (small molecule, biologics, cell or gene therapy) or more than one therapeutic area.
\n

\n

General requirements

\n

\n
• Master's or PhD, or a current PhD candidate, in Biology or a directly related field (molecular or cell biology, genetics, immunology, neuroscience, biochemistry, bioinformatics, or computational biology), completed in the U.S., Canada, Europe, or the UK. For this role, directly related fields also include chemistry, medicinal chemistry, pharmacology, pharmaceutical sciences, and toxicology.
\n
• 3+ years of hands on experience in your subfield (see Domain qualifications above). Time in an academic lab or research institute counts after undergraduate education.
\n
• Able to draw on your own real world experience and day to day workflows to craft scenarios that test whether an AI system can actually do the work.
\n
• Hands on practitioner: you currently do (or recently did) the bench or analysis work yourself at an individual contributor level, not solely in a managerial capacity.
\n
• Full professional or native level written and spoken English, with strong written communication. You can explain complex scientific reasoning clearly and concisely, and articulate why a result is wrong, not only that it is.
\n
• Comfort with ambiguity and attention to detail. You can orient in a new set of files and build an accurate, deep working picture of it quickly, especially when the science sits partly or wholly outside your own specialization. You verify what a document claims against the underlying data.
\n
• Capable of interpreting feedback, judging which parts of it are actually correct, and applying it without hand holding. When stuck, you look for the answer rather than waiting for one.
\n
• Ability to ramp quickly on unfamiliar work from written material and instructions alone, including where that material is incomplete (for example, writing grading rubrics for the first time).
\n
• General familiarity with AI and LLM tools. You have used models like Claude or ChatGPT in life sciences professional work and have the judgment to tell a well reasoned answer from a plausible sounding but incorrect one.
\n
• Baseline tech literacy: comfortable with cloud file tools (e.g., Google Workspace), managing browser profiles, downloading and installing desktop apps (e.g., Claude), and everyday file handling (e.g., converting between Excel and Google Sheets, zipping files for sharing).
\n
• Available at least 10 hours per week, with no weekly maximum. Consistent availability is valued and more hours are welcome.
\n
• Based in the United States, Canada, or the UK (Ireland and Australia may also be accepted).
\n

\n

Compensation and terms

\n

\n
• Pay: $40 - $125/hr USD. Paid via PayPal on a regular cadence.
\n
• Contract position. Fully remote. Minimum 10 hours per week with no weekly maximum.
\n
• Flexible scheduling. Multi day task timers let you spread work across days.
\n
• Access to Claude and ChatGPT is provided through the project; no personal subscription is required.
\n

\n

What to expect

\n

\n
• Apply through the posting link and create your account.
\n
• Complete the paid skills assessment (about 4 hours). It tests domain fit, careful reading, task design, rubric judgment and your response to feedback on a prompt. All work must be your own; submissions produced with AI tools are rejected, and this is the single most common reason candidates do not pass.
\n
• Our team reviews your assessment. If you pass, you complete onboarding and a short training project that walks through how tasks, files, and rubrics are built.
\n
• Propose a task from your professional experience: the scenario, the prompt, the files you will build or choose to use, and where you expect the model to fail. An expert reviewer reads it and sends written feedback either way.
\n
• Once your proposal is accepted, build the task in full and run it against frontier models. Every completed task goes through expert review, and experienced contributors are invited to review and refine other experts' tasks.
\n
• Each stage is a gate: work must be accepted before you move on. You are compensated for every step you complete, but not every submission is accepted. Feedback and revision are a normal part of the process. Support is available through platform instructions, onboarding materials, a dedicated Slack channel, and office hours.
\n

\n

About DataAnnotation

\n

DataAnnotation works with frontier AI labs building the world's most advanced models, with the goal of enabling human aligned AI. Our expert programs bring together practicing professionals to evaluate and improve what these models can do in their fields.

\n

DataAnnotation believes that human expertise is essential to building trustworthy AI. We don't want AI training AI. We want real scientists in the room. If you have a strong life sciences background and want to put it to work in a new way, we'd love to hear from you.

\n

#lifescience