Oxbridge Assessments

High-volume hiring funnels have a trust problem: you can test thousands of candidates cheaply, but the moment a test is remote, you have to catch outside help without a human watching every screen. Oxbridge is built to do both at once.

It's a complete platform — candidate assessment flow, admin marking and question-bank management, and a recruiter portal — sitting on 11 data models and 34 API endpoints. It was engineered to process a roughly 5,000-candidate pipeline, and it's launching.

The assessment engine

Candidates move through a five-part battery — reading, writing, speaking, multiple-choice and scenario. Each type has its own capture and marking path; admins manage the question bank through full CRUD and mark the open-ended sections.

Proctoring integrity

Because nobody is physically invigilating, integrity is inferred from signals gathered during the sitting:

• Tab-switch and focus-loss events — did the candidate leave the test?
• Keystroke-pattern analysis — does typing behaviour suggest pasted or dictated help?
• Webcam capture — periodic frames tied to the attempt.

Deterministic scoring

Every closed-form answer runs through a deterministic scoring engine: the same answers always yield the same score. That's not a detail — hiring decisions have to be defensible and auditable, and non-determinism would make a disputed result impossible to defend.

Security posture

Candidate PII is encrypted at rest with AES-256-GCM; credentials are bcrypt-hashed; endpoints are rate-limited. The threat model treats candidate data as sensitive and the scoring pipeline as something that must resist tampering.