Senior Applied Data Scientist, Learner Modeling

At Instructure, we believe in the power of people to grow and succeed throughout their lives. Our goal is to amplify that power by creating intuitive products that simplify learning and personal development, facilitate meaningful relationships, and inspire people to go further in their education and careers. We do this by giving smart, creative, passionate people opportunities to create awesome. And that's where you come in: Our team builds AI-native capabilities, reusable AI systems, and shared infrastructure that power multiple products and workflows across the platform. We are looking for a Senior Applied Data Scientist to develop and validate knowledge tracing and longitudinal learner models, defining what "mastery" means operationally and ensuring the outputs are trustworthy, calibrated, and fair before they reach learners and educators. This is a measurement role as much as a modeling role, closer to computational psychometrics than to generic data science: a prediction that is accurate on average but miscalibrated, unstable, or unfair is not product-ready, and judging that difference is the core of the job. You will partner with AI platform engineers to productionize training and scoring pipelines and to monitor quality in live environments. You will work closely with product, engineering, and research partners to turn advanced AI ideas into reliable product capabilities used at scale. Important note on scope: This role is judged on the validity of what the models claim about a learner (calibration, fairness, and stability over time), not on predictive accuracy alone, and not on BI/reporting or experimentation analytics. We are looking for someone who can define a construct, model it rigorously, and stand behind the result in a live product. WHAT YOU'LL DO - Design and build knowledge tracing and longitudinal learner models that support mastery and progression features surfaced to customers - Define what "mastery" and "progression" mean operationally: constructs, model targets, and evaluation criteria aligned to learning outcomes and product requirements - Build robust training and scoring approaches for noisy, incomplete, and evolving learner interaction data - Own model trustworthiness: lead evaluation for validity, calibration, fairness, stability over time, interpretability, and failure modes (not predictive accuracy alone), and set the bar for what is trustworthy enough to ship - Partner with engineering to productionize learner models into reliable services, including deployment, monitoring, and iteration loops - Collaborate with product and learning partners to translate learning theory into scalable product systems, and to communicate model behavior, assumptions, and limitations clearly WHAT YOU'LL NEED - 6+ years of experience in applied machine learning, data science, or applied research, with ownership of models shipped into real products - Strong depth in at least one of: knowledge tracing, sequence modeling, probabilistic modeling, temporal modeling, Bayesian approaches, or psychometrics / item response theory - A working understanding of computational psychometrics: the rigorous measurement of a construct, not only the optimization of a predictive metric - Demonstrated ability to evaluate a model on more than accuracy, including calibration, uncertainty, fairness, robustness, and interpretability, and to make a defensible ship/no-ship call on that basis - Experience working with longitudinal data and designing models that remain stable, meaningful, and interpretable over time - Strong Python and ML stack skills, with the ability to implement and iterate on modeling pipelines - Ability to communicate modeling choices, assumptions, and uncertainty clearly to technical and non-technical stakeholders IT WOULD BE A BONUS IF YOU HAD - Deep experience with computational psychometrics, psychometrics, educational measurement, or item response theory, with a track record of bringing that rigor into machine-learned m