Riiid Labs : Adaptive AI Engine for Personalized Learning and Test Mastery

Education / EdTech

Riiid Labs – AI-Powered Personalized Learning Platform

Riiid Labs partnered with us to develop an adaptive AI learning system that personalizes to individual student progress and competency. Our system analyzes individual learning behaviors, test performance, and engagement metrics to tailor content in real-time. Using reinforcement learning and predictive modeling, it identifies areas where students struggle and delivers hyper-focused practice. The AI also supports tutors with insights into learning gaps and knowledge retention patterns. This personalized approach helps improve outcomes in standardized test preparation. Riiid’s platform is now one of the most advanced AI-driven education tools globally.

Project Overview

Client: Zalando (Europe’s leading fashion platform, €10.3B revenue)
Limitations: Conventional learning platforms offered no adaptive instruction
Goal: Build AI fashion assistant to:
- Personalize outfit suggestions through natural conversations
- Reduce decision fatigue and browsing time
- Seamlessly plug into live inventory, trend APIs, and CRM
My Role: AI Lead Engineer (GenX Software Team)
Team: 9 (3 NLP Engineers, 2 Frontend Devs, 2 Data Scientists, 1 Trend Analyst, 1 PM)
Timeline: 5 months (Prototype → Multi-country rollout)

“Thanks to GenX, we don’t just deliver learning—we deliver outcomes. Our platform adapts to every student like a personal coach.”

Head of AI & Learning Science, Riiid Labs

The Challenge

Critical Pain Points:

Students had to follow a one-size-fits-all learning path, slowing progress
Educators lacked visibility into which concepts learners truly struggled with
High content drop-off rates due to disengagement and misalignment with mastery levels

Technical Hurdles:

Modeling learning behaviors and predicting knowledge retention over time
Aligning AI-generated lesson plans with curriculum standards and assessments
Building explainable reinforcement learning pipelines for transparency and trust

Tech Stack

Component	Technologies
Learning Models	RLlib (Ray), BKT, PyTorch, XGBoost
Data Engineering	AWS Glue, Redshift, Firehose
Analytics & Dashboards	Metabase, D3.js, Looker Studio
Backend + APIs	Node.js, FastAPI, PostgreSQL, GraphQL
Frontend + SDKs	React, Flutter SDK, WebRTC (for tutor sessions)
Monitoring & Privacy	Datadog, Sentry, FERPA & GDPR-compliant encryption

Key Innovations

AI tracked learning patterns and pinpointed weak topics using predictive modeling. Real-time adaptations boosted test readiness and reduced dropout rates. Tutors received detailed performance maps to tailor interventions.

AI-Curated Adaptive Pathways

Each student followed a unique trajectory optimized for mastery

Result: 38% faster topic completion compared to static curriculums

Predictive Intervention System

Identified early signs of dropout or concept fatigue

Result: 27% improvement in session completion rates

Mastery-Driven Tutor Dashboard

Gave educators real-time visibility into progress and gaps

Result: 31% boost in student performance after targeted review sessions

Our AI/ML Architecture

Core Models

Personalized Learning Engine:
- Reinforcement Learning model that optimizes question sequencing
- Bayesian Knowledge Tracing (BKT) to map mastery progression
Engagement Predictor & Intervention Module:
- Classifies students at risk of disengagement
- Triggers personalized nudges and gamified tasks to re-engage
Educator Insight Dashboard:
- Predictive heatmaps of student performance across topics
- Retention curve analysis and mastery decay detection

Data Pipeline

Sources
- Student responses, quiz/test scores, content completion data
- Engagement metrics (time-on-task, skips, reattempts)
- Tutor feedback and session notes