Learning Scientist (Education Intelligence)

Team: Education Intelligence — Learner Intelligence Modelling Reports to: [Head of Education Intelligence / Chief Digital & Product Officer] Location: [Remote / hybrid] Level: [Senior] About companytmrw education is creating an education-intelligence arm of TMRW, the AI-first platform for global education systems. We're turning fragmented school systems into one intelligent ecosystem — where AI proposes, humans decide, and evidence drives every action — live and expanding across the region. We bring together school operations and education insight in one place, so leaders and teachers can spend less time on admin, make better decisions, and focus on learning. Built for complex, global school networks, tmrw is committed to thoughtful innovation, user-centred design, and practical tools that work in real schools. About the role The Learning Scientist defines the educational meaning that turns raw school data into trustworthy intelligence. Sitting at the intersection of learning science and AI, you decide what a strength, a risk or a readiness signal actually is — and which data sets should be read together to support it — so the platform reasons about learners the way an expert educator would. You are the bridge between educational evidence and the machine. You define the constructs, the combinations and the confidence thresholds that let AI propose, humans decide, and evidence drive every action — in real schools, at scale What you'll do Define educational constructs — turn strengths, behaviours, engagement, wellbeing, growth and support needs into clear, valid definitions the platform can reason over. Decide which data belong together — determine which data sets should be considered as one signal — for example attendance, task completion and wellbeing combining into learning reliability — and which must never be conflated. Build the semantic learning model — map the school's many data sources into a shared set of meaning categories, so raw data becomes educational meaning rather than disconnected datasets. Design signal recipes — specify what evidence, and in what combination, defines each signal — strength, risk, growth, readiness, support need — leaving the scoring and calibration to Data Science. Set the rules for trust — define what kinds and weight of evidence make an inference defensible, and when a human must review — with Data Science calibrating the scores that implement it. Validate in real schools — pressure-test constructs and explanations with teachers, leaders and learners, so every recommendation makes sense to the people who act on it. Guard against misreading learners — work with Responsible AI and safeguarding colleagues to prevent bias, over-reach and false signals across GEMS and partner schools. Keep intelligence reusable — ensure the meaning you define is shared across Learner360 and every teacher, parent and student experience — built once, applied everywhere. How the role fits You'll sit at the centre of the Learner Intelligence Modelling squad, translating educational evidence into shared intelligence across three functions: Data Science & AI/ML — turn the constructs you define into scoring, pattern detection and model-based signals. Data & Ontology Architecture — shape your meaning categories into the canonical learner graph and its semantic relationships. Product, UX & Responsible AI — decide which signals matter first, and validate that explanations are clear, fair and safe. What you'll bring Real education experience — as a teacher, learning scientist, assessment specialist or educational researcher — with a deep feel for how learning shows up in school data. A grounding in learning science, educational measurement, psychometrics or cognitive science at graduate level, and the judgement to say which evidence supports a valid inference and which does not. The ability to decide which data sets belong together and which should stay apart — and to explain why in plain language. Comfort working alongside data scientists, engineers and ontology designers; you don't need to code, but you do need to shape what they build. Care for fairness, safeguarding and the limits of inference when the subject is a child. Clear communication — you can make a construct or a confidence threshold understandable to both a teacher and an engineer. Nice to have Experience designing rubrics, learning progressions or competency frameworks. Familiarity with school data sources — SIS, LMS, markbook, attendance, wellbeing and co-curricular systems. Exposure to knowledge graphs, ontologies or semantic models. Work across multiple curricula or international school networks.  What success looks like Raw school data becomes trusted educational meaning — the same construct means the same thing in every experience. Signals are evidence-based and explainable, so teachers and leaders act on them with confidence. New agents consume the shared meaning you defined instead of rebuilding intelligence each time. Recommendations about learners are fair, safe and defensible in real schools.