Overview
Key Features
- AI-driven Scoring: Uses OpenAI models for automated scoring of speaking and writing tasks with explainable feedback and improvement tips.
- Multitenant Architecture: Tenant isolation, configurable exam templates, and per-tenant analytics for schools and partners.
- Real-Time Proctoring & Analysis: Live audio/video capture, behavior analysis, and integrity checks during exams.
- Payment & Billing: Integrated with Stripe for subscription and one-time payments, per-tenant billing and coupon support.
- Notifications & Email: Resend integration for transactional emails (receipts, results, reminders).
- Practice Modes & Mock Tests: Full-length timed mocks, section-level practice, and performance dashboards.
- Admin Dashboard: Manage tenants, exams, candidates, and review AI-generated scores with human override.
Technologies Used
- Frontend: React for the candidate and admin interfaces
- AI: OpenAI for scoring, feedback, and conversational practice
- Backend: Express/Node.js for APIs and orchestration
- Database: Supabase for multitenant data models and real-time features
- Payments: Stripe for subscriptions and payments
- Email: Resend for transactional email delivery
- Storage: Cloud object storage for recordings and resources
Challenges and Learnings
- Fair and Explainable Scoring: Designing prompts and evaluation criteria to provide consistent, unbiased scoring and actionable feedback.
- Tenant Isolation: Ensuring data separation while allowing centralized administration and analytics.
- Low-Latency Recording Processing: Efficient pipelines for processing audio/video segments for scoring and proctoring.
- Regulatory Compliance: Handling PII, recordings, and payment data with proper encryption and retention policies.
Outcome
Impact
- Deployed as a multi-tenant SaaS for language schools and test-prep providers
- Scaled to thousands of mock exams per month with secure recordings
- Reduced manual grading time by over 70% through AI-assisted scoring
Try it
React
OpenAI
Supabase
Stripe
Node.js
Resend
