Loyola Tutoring
Messaging, access control, and AI-assisted learning across a connected tutoring platform.
Overview
As an independent software developer for Loyola Tutoring, I contributed to a multi-app platform spanning administration, parents, tutors, and students. My work focused on messaging, secure integration with the existing practice site, and reliable AI-assisted learning workflows.
Contribution Software Developer — Independent Contractor
Problem
The new enrolment system and legacy practice site needed to work together without disrupting access. Staff and parent notifications needed reliable delivery and traceability, while learning workflows needed resilience when an AI provider became unavailable.
Approach
- 01
Built a transactional Postgres outbox with scheduled dispatch and Deno edge functions for Twilio SMS and Postmark email, including signed delivery webhooks, editable templates, a compose interface, and message logs.
- 02
Connected the new enrolment database to the legacy Prisma and Neon practice app through an authenticated access API with paid-booking rules and outage-safe fallbacks.
- 03
Routed essay-marking and extraction calls through OpenRouter with fallback models, strict output schemas, and zero-data-retention configuration.
- 04
Hardened multi-role data access with scoped database functions and row-level security, backed by rollback-safe SQL audits and authorization checks. Also contributed parent, tutor, and student-facing features.
Outcome
- Shipped the production email and SMS layer supporting 13 automated staff and parent alert types.
- Launched the enrolment-to-practice access gate for roughly 400 family accounts at the term cutover.
- Added provider resilience to AI marking and extraction through fallback models and schema-validated responses.
- Implemented and audited access controls across multiple user roles, with six new edge functions and 29 Postgres migrations documented in the work audit.