An examination platform for society recruitment where delivering the questions was the easy half. Stopping people cheating was the project.

01 Architecture

• Django REST on PostgreSQL with a React client, plus a separate face recognition service in Python.
• Tab switch detection with three strike audit trails written server side, not trusted to the client.
• Webcam capture every 60 seconds through the exam window, stored against the attempt.
• Randomised question draws per candidate, so two people sitting together see different papers.
• Face service verified candidates against registered photos, handling 936 encodings across 2,000 candidates.

02 What is in it

• Ran recruitment for the society at full intake scale.
• Every anti cheat signal is auditable after the fact, not just enforced live.
• Identity verification separated into its own service so the exam path stays fast.