DevPost — About The Project (Short Version)

Inspiration

WIAL certifies Action Learning Coaches globally, but each chapter ran its own website independently — no shared coach directory, no unified certification tooling, no way for organizations to easily find certified professionals. We wanted to fix that.

What it does

A multi-tenant platform for WIAL's global chapters:

Coach directory — AI-powered search (natural language) across certified coaches worldwide
Certification hub — guides coaches through CALC → PALC → SALC → MALC with documents, LMS links, and Credly badges
Chapter microsites — each WIAL chapter gets its own branded subdomain
Admin tools — coach approval workflow, content management, role-based dashboards
AI chatbot — answers WIAL certification questions site-wide
Payments — dues collection via Stripe
Audio narration — ElevenLabs TTS for content pages and coach profiles

How we built it

Next.js 15 (App Router) + TypeScript + Tailwind CSS v4
Supabase (PostgreSQL + pgvector + Row-Level Security) for database and auth
Multi-tenancy via Next.js Middleware subdomain rewrites
AI search — Cohere embeddings + pgvector + Claude 3.5 Haiku (query parsing) + keyword fallback
Chatbot — GPT-4o-mini via OpenRouter
Audio — ElevenLabs TTS stored in Supabase Storage
Payments — Stripe Checkout with webhook handling
Email — Resend for coach approval/rejection notifications

Challenges we ran into

Getting multi-tenant routing clean in Next.js without tangling the codebase
AI search quality — vector search alone wasn't enough; layering in keyword fallback and LLM query pre-processing made it actually useful
Designing a certification schema flexible enough for WIAL's rules while keeping RLS policies correct across all five roles
Triggering embedding regeneration when a coach is approved without race conditions

Accomplishments that we're proud of

Full multi-tenant architecture deployable as a single Vercel project
Hybrid AI search that's genuinely useful, not just a demo
End-to-end coach lifecycle: register → approve → search → certify → pay
Built-in audio accessibility from day one

What we learned

Multi-tenancy has to be a day-one architecture decision
Good AI search is layered: embeddings + keywords + LLM intent parsing
Supabase RLS policies are code — treat them that way
Next.js 15 Server Actions + Supabase is a clean, secure pairing that eliminates a lot of API boilerplate