Adaptive Learning AI for an EdTech Platform
The Challenge
The platform's curriculum was entirely fixed: every student followed the same sequence at the same pace. High-performing students found the pace too slow and disengaged. Struggling students fell behind without the additional support they needed and dropped out. The churn rate was high and outcomes were poor across both groups — a structural problem that content improvements alone couldn't solve.
Our Approach
Tequity built an adaptive learning engine that models each student's knowledge state continuously using Bayesian knowledge tracing — a well-validated approach from educational psychology. As students complete exercises and assessments, the system updates their mastery estimates for each concept in the knowledge graph and selects the next piece of content most likely to advance their learning efficiently.
Content delivery adapts along three dimensions: topic sequence, difficulty level, and format (video, practice problem, explanation, worked example). The system also identifies at-risk students early — those whose engagement or performance patterns predict dropout — and triggers targeted interventions.
The Results
Learning outcomes improved 40% versus the fixed-curriculum control group in an A/B test run across 10,000 students. Completion rates increased 65% as learners received appropriately challenging content rather than pacing out or getting lost. The personalized path system now serves 50,000+ active learners, with zero additional infrastructure cost per student.
