Case Study

ALI

Adaptive Learning Intelligence — an AI-powered platform that personalises education through real-time learner analytics, intelligent content sequencing, and machine-learning-driven path optimisation.

Discuss this project →
Overview

The challenge

Traditional e-learning platforms deliver the same content sequence to every learner, regardless of prior knowledge, learning pace, or cognitive preferences. The result is disengagement: advanced learners are bored, struggling learners fall behind, and completion rates hover around 10-15%.

ALI was conceived to solve this by treating each learner journey as a unique, dynamically evolving path. Rather than static curricula, ALI builds a real-time model of learner mastery and adapts content difficulty, format, and sequencing on the fly.

3.2x
Improvement in completion rates
47%
Faster time-to-mastery
89%
Learner satisfaction score
12k+
Active learner profiles
Core capabilities

What ALI does differently

Adaptive Path Engine

A Bayesian knowledge-tracing model continuously estimates learner mastery across skill nodes. Content difficulty auto-adjusts within a zone of proximal development.

Multi-Modal Content Delivery

ALI detects engagement patterns to determine whether a learner responds better to video, interactive simulations, text, or problem sets, then re-weights the content mix in real time.

Predictive Analytics Dashboard

Instructors get early-warning flags for at-risk learners, cohort heatmaps showing concept-level gaps, and AI-generated intervention suggestions.

Spaced Repetition

An SM-2+ scheduling algorithm surfaces review prompts at scientifically optimal intervals, reinforcing long-term retention without manual study planning.

LLM-Powered Tutoring

Integrated large language model provides Socratic dialogue, contextual hints, and worked-example generation, adapting its teaching style to the learner mastery level.

Gamification Layer

Streak tracking, skill trees, challenge modes, and peer leaderboards drive intrinsic motivation while feeding behavioural data back into the adaptation engine.

Journey

Project timeline

Q3 2024 - Research

Literature review and prototyping

Surveyed adaptive learning frameworks (IRT, BKT, DINA). Built proof-of-concept with synthetic learner data. Validated Bayesian knowledge tracing outperformed static prerequisite chains.

Q4 2024 - MVP

Core engine and pilot launch

Deployed adaptive engine on Cloud Run. Integrated with a pilot cohort of 200 learners across two university courses. Collected baseline metrics for A/B comparison.

Q1 2025 - Scale

Multi-tenant architecture and LLM integration

Refactored for multi-tenancy. Added LLM tutoring layer with Claude API. Built instructor dashboard with predictive at-risk alerts. Scaled to 5,000+ active learners.

Q2 2025 - Present

Advanced analytics and research publication

Introduced cohort-level learning analytics, A/B experiment framework, and gamification engine. Currently preparing research paper on adaptive path efficacy.

Technology

Tech stack

ALI is built with a modern, cloud-native stack optimised for real-time responsiveness and horizontal scaling.

ReactNext.jsTypeScriptTailwind CSSFastAPIPythonPyTorchClaude APICloud RunFirestoreBigQueryPub/SubRedisFirebase AuthTerraformGitHub ActionsWebSocketsDocker
My role

What I built

As sole architect and developer, I designed and implemented every layer of ALI, from the Bayesian adaptive engine and LLM tutoring integration to the React frontend, instructor analytics dashboard, and GCP infrastructure.

This project sits at the intersection of my dual expertise: 12 years of marketing technology leadership (understanding how to drive engagement and retention) combined with ongoing doctoral research in Educational Technology at the University of Cyprus (grounding every design decision in learning science).

Key contributions include the knowledge-tracing algorithm (adapted from BKT with novel extensions for multi-skill dependencies), the real-time content-sequencing pipeline, and the predictive at-risk model that gives instructors 2-3 weeks of early warning before learner dropout.

Interested in ALI?

Whether you are an institution exploring adaptive learning, a researcher in EdTech, or a company rethinking corporate training, I would love to talk.

Get in touch →