Back to Projects
E-learning, Medical Education

Medify

Built three AI systems for a medical-education platform that students use to revise against exam questions — a content audit engine, a knowledge-graph recommendation engine and a conversational study assistant.

Medify
UK

Key Results

~95%

Content-audit accuracy across the testing dataset

~80%

Reduction in the content team's manual effort

3

AI systems delivered (audit, recommendation, assistant)

The Challenge

Medical students revise by working through thousands of exam-style questions, so the quality of that question corpus is everything. But as the dataset grew, overlaps, inaccuracies and gaps crept in — and finding them meant the content team manually reviewing huge volumes of questions, a slow and low-ROI task.

At the same time, every student was served broadly the same content, with no intelligent sense of what each individual should revise next.

The Solution

Acting as fractional Lead AI & Full-Stack Engineer, we designed and built three complementary AI systems, using Python, scikit-learn, TensorFlow, AWS and React.

AI Content-Audit Engine

Audited the question corpus to surface overlaps, inaccuracies and white space across the testing dataset with ~95% accuracy.

Knowledge-Graph Recommender

Turned each student's progress into a hyper-personalised, next-best learning path.

Conversational Study Assistant

Let students ask questions about the curriculum, individual test items and what to revise next.

The Impact

Tangible Outcomes

AI content audit reached ~95% accuracy across the testing dataset

Cut the content team's manual review effort by ~80%

Delivered a hyper-personalised, knowledge-graph learning-path engine

Launched a conversational AI study assistant for students

Key Takeaway

Embedded as the lead AI and full-stack engineer to take three distinct AI systems from concept to production, removing a major content bottleneck while making revision genuinely personalised.

Want to Put AI at the Heart of Your Learning Platform?

Let's discuss how we can build AI systems that personalise learning and remove manual bottlenecks.