iOS.
iOS app @sushrutalgs.ai
Built the native iOS app for sushrutalgs.ai, an AI study companion that answers surgical-exam questions with streaming responses backed by citations, figures, and tables from standard textbooks. It ships three sign-in options, conversation history that syncs across a user's devices, and an iPhone and iPad interface.
Source is private; sushrutalgs.ai is a live product. Happy to walk through the code or grant read access on request.
Stack
- Swift 6
- SwiftUI
- supabase-swift
- Google Sign-In
- Sign in with Apple
- Xcode Cloud
System architecture. Tap to enlarge.
Overview
The iOS app is the native client for sushrutalgs.ai. It gives surgery residents a fast, polished way to study on iPhone and iPad: ask a question, watch the answer stream in with citations, figures, and tables, and pick up the same conversation on another device.
Approach
- Native and modern. SwiftUI only, iOS 26, Swift 6 strict concurrency, with a Model-View architecture and observable services injected at the root.
- Streaming chat. A five-type server-sent-events protocol drives a per-bubble typewriter reveal, with cross-device handoff broadcast over a Supabase Realtime channel.
- Secure auth. Email one-time codes, Google Sign-In, and Sign in with Apple, all exchanged through Supabase with nonce-based replay protection.
Results
About 33,200 lines of Swift across 105 files and 80 views, shipping at a 20.8 MB App Store install size. A CI security gate fails any build that would leak a backend host or secret, and the conversation model and streaming protocol are shared one-to-one with the web client.
Engineering
Swift 6 and SwiftUI with Swift Package Manager, supabase-swift for auth, Realtime, and database access, GoogleSignIn and Sign in with Apple for identity, and Xcode Cloud for CI. It is a leaf client: all model inference goes through the sushrutalgs.ai BFF, so it holds no backend secrets.