Required experience and skills

• 5+ years building predictive models that someone actually used, including a few years where you owned the problem rather than being handed it, with the judgment that goes with it: calibration, threshold-setting, and knowing when a feature is leaking the answer.

• Practical NLP experience: text classification, clustering, embeddings, or similar applied work. Calling an LLM API is useful but isn't the same thing.

• Strong SQL as a primary tool, not a way to get data into a notebook.

• Working Python for modeling and analysis.

• Solid applied statistics, with the judgment to know which method fits the question and when the data can't support a conclusion.

• Modern cloud warehouse experience (Snowflake, BigQuery, Databricks, or similar), especially with in-warehouse AI or agent tooling.

• Excellent written and verbal communication. You can hand a finding to a non-technical colleague and have them act on it, including knowing what would change your conclusion.

• Care about how predictions get used. Our scores influence how students get supported, so we want someone who checks whether a model works as well for part-time students as for everyone else, and says so when it doesn't.

• Comfort with ambiguity and honesty about uncertainty. We'd rather hear "the data can't answer this" than a confident answer that falls apart later.

• Comfortable working in version control with code review, so your analysis and models are reproducible by someone else.

• Experience productionizing model output into an operational workflow.

• Experience scheduling and monitoring recurring jobs in production, and the judgement to reach for tooling the team can maintain rather than a specialized stack that only you know.

• Track record of choosing what to work on. You’ve turned an ambiguous business goal into a scoped project, made the prioritization case, and been accountable for whether it mattered.

Nice to have

• Transformation tooling (dbt or similar), dimensional modeling, or analytics engineering exposure.

• Familiarity with AI evals or prompt evaluation.

• A modern BI tool (Sigma, Looker, Hex, Tableau, or similar) for making findings usable by others.

• Experience working closely with analytics or data engineers, where your models depended on someone else's tables.

• Linguistics or computational linguistics background for intent classification and evaluation work.

• EdTech, higher education, or student success background.

This probably isn’t the right opportunity for you if

• You need a dedicated ML platform to be effective. We deliberately don't run one: models run inside our cloud warehouse, features come from our transformation layer, predictions land in tables that downstream systems read.

• Your experience is primarily research or model development without anyone using the output.

• You want to specialize. The role ranges across modeling, text analysis, evaluation, and enablement, and none of them will be someone else’s job.

• You’d rather have your work reviewed rather than review others. As the senior person in this discipline here, you’ll be setting the standard, not inheriting one.

• You’d introduce a new tool for every problem. We optimize for delivering business value using long-term maintainable solutions, which sometimes means the second-best tool.

Key Performance Metrics

• Validated predictive signals shipped and in active use by partners or internal teams.

• Model quality in production terms: calibration and precision at an actionable alert volume.

• Adoption of AI data tooling, including share of questions answered without data team involvement.

• Evaluation coverage: share of data team AI surfaces with an active eval, and whether model or prompt changes ship with evidence.

• Stakeholder feedback from Partner Success, Product, and Leadership.

• Quality of prioritization: whether the work you chose turned out to matter, and whether you surfaced tradeoffs early.