Token-streamed AI chat client on iOS

Fitness & Biomechanics

Industry: Fitness & Biomechanics

Platform: iOS

Stack: Swift · SwiftUI · Vision · AsyncSequence · Combine

Description: Native iOS AI-chat client with SSE token streaming, on-device OCR, and strict-concurrency safety.

---

A wellness platform required a native iOS client built against an existing LLM-backed REST contract. The engineering surface was a real-time chat over Server-Sent Events with token-level rendering, an on-device OCR pipeline that extracts text from screenshots before any server contact, and full Swift strict-concurrency compliance across 16 API endpoints. Callback-based system frameworks needed to bridge cleanly into actor-isolated state without runtime data-race crashes.

We owned the full iOS client — architecture, networking, OCR pipeline, navigation, and the contract-driven test layer. The client owned the LLM backend, design system, and 16-endpoint contract we built against.

Engagement: ~8-week build · single iOS specialist · 2026.

Engineering signal: zero Swift strict-concurrency warnings across 156 protocol-driven tests; SSE token rendering with no main-thread blocking.