A question bank that studies the student

Practice sessions now aim at each student's weakest topics automatically — the platform does the studying about the student, so the student can study the material.

Context

USMLEPrep serves medical students preparing for the United States Medical Licensing Examination — one of the highest-stakes test series in any profession. Study time is the scarcest resource these students have.

The problem

Generic question banks treat every student the same: same order, same mix, same review schedule. Students end up re-answering what they already know because it feels productive, while their actual weak areas get the leftovers.

What Haven built

• A practice engine that tracks per-topic performance and weights upcoming sessions toward weak areas
• Spaced repetition for missed questions, scheduled against the student's exam date
• Explanations that connect each answer to the underlying concept, not just the correct letter
• A progress view built around one question: "what should I do today?"
• FERPA-aware data handling for institutional partnerships

Results

Students spend their sessions on the material that needs them. Session composition, review scheduling, and progress reporting all run from the same per-topic performance model — one source of truth instead of three features that disagree.