
Software Engineer
Audria · Nov 2025 – Present
- Shipped v1.0 to the App Store from scratch an always-listening, proactive personal AI assistant for iOS and macOS built with SwiftUI, AVFoundation, CoreML, and real-time speech pipelines.
- Architected the app's UI and data layer using MVVM, an offline-first repository pattern (SwiftData, cache-then-network synchronization, background replay), and a reusable design system, enabling scalable development across 20+ feature modules.
- Optimized SwiftUI performance using Instruments, reducing unnecessary view invalidations, improving responsiveness, and lowering memory and CPU overhead.
- Replaced a SceneKit visualization with a custom Metal renderer featuring demand-driven rendering for improved performance and reduced idle resource usage.
- Shipped an on-device speaker diarization pipeline in Swift (pyannote-3.0 segmentation TitaNet speaker embeddings NME-SC clustering) running fully offline with no audio leaving the device for speaker identification and diarization.
- Built a real-time AVAudioEngine capture pipeline with streaming VAD, wake-word detection, acoustic echo cancellation and vDSP-accelerated PCM conversion; eliminated an audio-loss class of bugs on stop by awaiting in-flight uploads under a cancellation shield and switching to capture-time timestamps.
- Architected an iOS BLE SDK for a custom wearable using L2CAP CoC for LC3 audio streaming, secure device authentication (RSA-2048-PSS + AES-128-CMAC), and a macOS peripheral simulator for parallel firmware development.



