Project

Riverside Sec Oral Conversation

An AI oral-exam grader for thirteen SEAB levels across three languages, with the teacher as the final authority on every mark.

AI oral-exam grading for thirteen SEAB exam levels across English, Chinese and Malay. Students record in the browser; the system transcribes, grades against the rubric, and returns feedback in minutes. The teacher remains the final authority on every published mark.

The Riverside Sec Oral Conversation sign-in page
The platform students and teachers sign in to: record an oral response, get SEAB-aligned grading in minutes.
A student recording an oral response in the browser
A student records their planned response and spoken interaction against the set task and its stimulus.
An AI-graded oral response with banded scores
The AI grades the response against the SEAB rubric and returns banded scores in about twenty-four seconds.
The AI's per-criterion written feedback and a short how-to-improve
Written feedback on each criterion explaining the marks, with a short 'how to improve' for the student to act on. Every mark is teacher-reviewed before it stands.
Admin analytics of usage across the school
Running across the school: thousands of practice recordings graded across thirteen SEAB levels in three languages.
Role
Solo: pedagogy, design, build, and operations
Stack
Cloudflare Pages, Claude (grading), Gemini (transcription), SEAB rubrics, Browser audio capture
Outcomes
  • Thirteen SEAB exam levels across English, Chinese, and Malay
  • 37 teachers and 700+ students at Riverside Sec
  • Around 8,000 practice recordings graded
  • Rubric-anchored feedback in about 24 seconds, with the teacher final on every mark
Read the essay → Visit the live app ↗ Live at Riverside Sec (staff & students)