Predrag Jevtic

I built a portable configuration layer that lets a single set of agent definitions and reusable skills run across multiple AI coding editors, rather than being rewritten for each vendor's format. Every editor ships its own convention for agent instructions, tool access, and project context, which forces teams to maintain the same knowledge three or four times and lets the copies drift apart.

I designed a shared source of truth for skills, agent roles, and tool configuration, then mapped it into each editor's expected structure so the same workflow behaves consistently regardless of which client an engineer opens. I structured skills as composable units with explicit trigger conditions, scoped tool permissions, and versioned instructions so that a workflow can be reviewed and improved like any other code artifact.

Finally, I documented the system and published it so other engineers can adopt the setup without rebuilding it.

I rewrote the AMD Link iOS client from UIKit to SwiftUI as one of two iOS engineers, working alongside a two-person Android team to hold feature parity across platforms. AMD Link streams gameplay and desktop content from an AMD Radeon-powered PC to iPhone, iPad, and Apple TV, so the rewrite had to preserve low-latency video playback and real-time GPU telemetry while the entire view layer was replaced.

The project involved rebuilding navigation, state management, and view architecture using SwiftUI and Combine, and migrating streaming controls, performance metric displays, hotkey configuration, and adaptive connection quality handling that adjusts stream settings based on available bandwidth. I reduced the view layer code and made the client faster to extend across iPhone, iPad, and tvOS targets. I also coordinated release timing, shared behavior with the Android team, and shipped the modernized client through App Store review as part of a four-engineer mobile group.

Architected and built the platform for a private aviation marketplace as its technical co-founder, writing the mobile client and the API myself. The product connects members booking private flights with charter operators, so the system spans a consumer app, operator tooling, payments, and live flight data.

Built the mobile client in React Native and Expo using continuous native generation, including native iOS Live Activities, so that members can follow flight status from the lock screen. Built the back end as a Swift Vapor API with BullMQ background jobs, integrating Stripe Connect and ACH via Plaid for payments, Twilio call masking to protect both sides of the marketplace, and flight tracking for live itineraries.

Designed the conversational AI layer: separate assistants for operators and for members, plus automated moderation of cross-party messaging. Built a knowledge graph service that exposes company and market context to AI tooling over MCP.