AI-powered recruitment ecosystem with 10 applications, 14 shared packages, and ~600K lines of code. End-to-end hiring pipeline for candidates, recruiters, and administrators.
A large-scale, AI-powered recruitment ecosystem built as a Turborepo monorepo with 10 applications and 14 shared packages. The platform serves three primary user surfaces: candidates build and refine their CVs with a personal AI career assistant; recruiters post openings and manage applications and hiring pipelines through a dedicated suite; and administrators access advanced analytics, feature flags, and system controls. These core surfaces are complemented by a public job board, a marketing site with a guest funnel, a support console, a headless API gateway, a content studio, and a technical documentation app.
The AI layer is built on Google Gemini 2.0 and the Vercel AI SDK with custom prompt engineering. It provides sector-agnostic ATS compatibility analysis and scoring, AI-generated cover letters, embedding-based semantic job–candidate matching, CV content suggestions, and bio enhancement. Candidates reach all of these capabilities through a conversational career assistant.
At its core is a strict layered separation: render-only apps talk to a central server layer through thin, business-logic-free tRPC routers; that server owns every query, mutation, and piece of domain logic via a typed context and dependency-injection pattern — over 300 procedures across nearly 30 domains. End-to-end type safety runs from database types all the way to client hooks via tRPC v11 and Zod. Asynchronous workloads are handled by dozens of typed background pipelines (onboarding, billing, notifications, job alerts, scheduled cleanup, and more) dispatched through a compile-time-checked registry.
The data layer runs on Supabase (PostgreSQL) with strict Row Level Security on every table, pgvector for semantic search, file storage, and realtime subscriptions. A subscription-based entitlement system works alongside PayTR payment integration. Server-side, multi-template PDF rendering for CVs, cover letters, and invoices — together with transactional email, realtime in-app notifications, structured logging, and Sentry observability — rounds out the supporting infrastructure.
The platform offers full internationalization for Turkish and English with RTL-ready layouts, and includes KVKK/GDPR-compliant consent management for analytics. A Radix-based shared component library on Tailwind CSS 4 keeps the interface consistent across every app. The fully typed, ~600K-line codebase is managed with Bun and Turborepo and deployed on Vercel.