Assessment Scientist to Upgrade AI Hiring Prompts (Fast Start)

We are Pynea, a London-based hiring intelligence platform. We replace CV-led screening with observable evidence of how candidates think, decide and adapt. Our system runs three assessment signals and produces a Living Profile of each candidate. We need someone to take our existing library of assessment prompts and rubrics and make them world class: sharper at eliciting real signal, tighter at scoring it, and defensible under UK and EU employment law. We move at pace. This has a fixed deadline. We need someone who can start within days and work in tight iterative loops with our team, not someone who disappears for a fortnight and returns with a document. This is not a prompt engineering job. Prompt craft is the surface. The actual work is assessment design: construct clarity, elicitation quality, rubric reliability, and adverse impact risk. If your background is LLM optimisation without assessment or selection experience, this will not be a fit. What you will do - Audit our current prompts and scoring rubrics against assessment design principles - Rewrite them for construct validity, consistency and reduced proxy signal for protected characteristics - Strengthen rubrics so two independent raters would reach the same score - Make every output explainable in plain language to the candidate it describes - Document your reasoning so our permanent team can maintain and extend the work What we need you to bring - Real experience designing or validating selection assessments, structured interviews or work samples - Working knowledge of Equality Act 2010 indirect discrimination risk, and ideally EU AI Act Annex III obligations for hiring systems - Practical fluency with LLM behaviour: you can write, test and iterate prompts yourself - Judgement about what makes a question reveal character rather than rehearsal Nice to have - BPS Chartered Occupational Psychologist, SIOP membership, or equivalent - Experience at an assessment vendor such as SHL, Arctic Shores, Applied, Sova, Talogy or HireVue - Published work on validity, fairness or adverse impact in AI assisted hiring How this runs - Full context and materials provided under NDA. - Shortlisted candidates complete a short paid work sample before we award the main contract. Strong likelihood of ongoing work beyond the initial scope. To apply Skip the generic cover letter. Answer the screening questions below. We read those first and nothing else. Tell us your earliest start date in your first line. Screening questions: 1. Describe one selection assessment you designed or validated. What construct was it measuring, and how did you know it worked? 2. A hiring prompt asks candidates to describe a time they overcame adversity. Name two ways that question could produce indirect discrimination risk, and how you would fix it. 3. We measure a construct that appears in no published competency framework. How would you validate it? 4. What is the most common mistake you see in LLM based candidate scoring?