Every feature is purpose-built around a single goal: closing the information gap between learner disengagement and institutional response.
Graph attention networks trained across EdNet, MoocCubeX, and OULAD datasets. 78% test AUC. Identifies at-risk learners 4–6 weeks before dropout — with explainable per-signal reasoning for every instructor.
AI generates targeted remedial paths in under 90 seconds when a learner underperforms. Adapts to pace, learning style, and engagement history. Goal-based, remedial, and skill-gap path types all supported.
Serve 12+ institutions from a single platform with full namespace isolation, per-tenant branding, and 4-tier role access (learner, instructor, admin, super-admin). No data bleed between tenants.
Per-module engagement rates, difficulty scores, and completion trends updated every 15 minutes. No separate BI tooling needed. Identify bottleneck items before they become dropout accelerators.
FNR gap monitoring keeps false-negative rate differences below 5 percentage points across demographic strata. Bias detection runs on every model retrain. Quarterly fairness audit reports are published publicly.
Multiple-choice, short answer, file upload, peer review, and proctoring-ready tracking with per-question item analysis. Attempt history, score trends, and mastery tracking per learner per skill.
SomaAI is designed around a closed loop: ingest data, surface risk, trigger action. Each step is automated; each alert is explainable.
Connect your existing courses and learner data via LTI 1.3 or xAPI. SomaAI establishes risk baselines and cohort profiles within 24 hours of connection.
The GAT model scores every learner daily across 127 behavioural signals. Instructors see at-risk flags weeks before they become dropouts — with the exact signals that drove each score.
Personalised paths, targeted nudges, and escalation alerts close the attainment gap. Real outcomes tracked per cohort. Weekly improvement reports delivered automatically.
SomaAI cut our second-semester dropout rate by 23% in the first cohort. The at-risk dashboard alone saved our intervention team 40 hours a week — time they now spend actually talking to students.
30-minute demo. No commitment. Deployment feasibility assessment included.