The problem
Competitive players lose matches to network problems they can’t control: bad ISP routing, jitter and packet loss between them and the game server. Products that fix this exist, but they are closed, expensive and often weak in some regions. The aim was to build a competitor from scratch: a packet-level engine, a relay network, and the commercial platform around it.
What I built
A packet engine in Rust. The engine intercepts only the target game’s traffic using WinDivert/Wintun and per-game IP filters. It tunnels that traffic to the nearest relay and injects the replies back transparently, so the game never knows it’s being routed. Sessions use X25519 key exchange with Ed25519-signed session tokens that can be verified offline. RFC 6479-style replay protection rejects duplicated or replayed packets.
A relay fleet. Lightweight relay servers in several data centers. GitHub Actions deploys them, so a fix to the relay code reaches every server without anyone logging in by hand.
A desktop app (Tauri 2 + React 19). It shows live ping, jitter and packet-loss tiles, a games grid and one-click routing. It updates itself from Cloudflare R2, and every release is signed so the updater can’t be hijacked.
A SaaS backend (NestJS 11 + Prisma + PostgreSQL).
- JWT authentication and license keys
- Stripe billing and an affiliate system
- Discord bot integration with role sync for the community
- Telemetry, audit logs and an admin API
A web platform (Next.js 16). Landing and pricing pages in English, French and Arabic (full right-to-left support), a customer portal, and a role-gated admin console.
Engineering decisions worth noting
- Only the game’s traffic goes through the tunnel. A per-game IP filter keeps browsers and downloads on the normal connection. Users never notice a VPN-style slowdown, and relay bandwidth costs stay predictable.
- Session tokens are checked offline. Relays verify signed tokens without calling the API, so an API outage never drops a live match.
- The release pipeline signs updates. Auto-updates are a supply-chain risk. Every build is signed, and the private key never touches CI logs or the repository.
- One monorepo. Turborepo + pnpm hold the shared types and configuration, so the desktop app, API and web portal never fall out of sync.
What this proves for your project
If you need low-level systems work (networking, performance, native apps) and the commercial layer around it (auth, billing, admin, deployment), this project covers the whole stack, built and run by one engineer.