Selected challenges

Magnosco

Integrating medical devices over Wi-Fi

Connected a dermatological imaging device to iOS over a device-hosted Wi-Fi network with no internet path, handling flaky handshakes, firmware version drift, and the need for a connection that a clinician could trust during a live examination.

Magnosco

Deploying multiple CoreML models in production

Shipped and version-managed several CoreML models on-device, each with different input pipelines and performance envelopes, while keeping app size, thermal impact, and inference latency within limits acceptable for clinical use.

Magnosco

Large-scale on-device AI optimization

Redesigned the execution flow of a memory-heavy computer vision workflow on iPad — introducing actor-based concurrency to coordinate pipeline stages safely, an intelligent caching strategy to avoid recomputing unchanged inference steps, and conditional model execution so expensive models only run when their output is actually needed.

Reduced peak memory usage from roughly 4GB to 1.2GB and eliminated the device crashes that came with it.

VNGRS · Diconium

Migrating Objective-C to Swift

Moved large, actively-shipping codebases from Objective-C to Swift incrementally, module by module, without freezing feature work or introducing regressions — using interop bridging and characterization tests as a safety net.

Diconium · Oculavis

Scaling modular Swift packages

Broke monolithic app targets into local Swift packages organized around features and layers, cutting incremental build times dramatically and letting teams own and test their modules independently.

Nord Security

Building encrypted cloud storage

Implemented client-side AES-256 encryption for a cross-platform cloud storage product, ensuring files were encrypted before they ever left the device and that the platform itself never held a usable key.

Momena

Privacy-first consumer application, owned end to end

Took Momena from concept to the App Store as a team of one — product design, architecture, development, security, and release — where privacy was the starting constraint, not a compliance checkbox: minimal data collection, on-device processing by default, and plain-language explanations for every permission requested.

Momena

StoreKit 2 subscriptions

Built a subscription system on StoreKit 2 end-to-end — entitlements, trials, family sharing, and receipt validation — designed to fail safe so a network hiccup never locks a paying user out of their own data.

Magnosco · Oculavis

Computer vision pipelines

Built OpenCV-based image preprocessing and feature extraction pipelines that turn raw camera or device sensor output into clean, model-ready input, tuned to run in real time on-device.

Magnosco

Medical imaging

Worked alongside scientists translating dermatological imaging research into a production pipeline, balancing image fidelity, patient privacy, and the accuracy requirements of a diagnostic support tool.

Magnosco

Secure medical device communication

Designed and implemented mutual TLS between an iPad application and a medical imaging device, including full certificate lifecycle management — provisioning, rotation, and revocation on-device — so every request was authenticated in both directions, not just the client's.

VNGRS

Large-scale SwiftUI migration

Led the incremental migration of a high-traffic streaming app's UI layer from UIKit to SwiftUI, screen by screen, keeping performance and App Store review timelines intact for a product used by millions.