███████╗██████╗  ██████╗ ██╗    ██╗██╗    ██╗ █████╗     ██╗      █████╗ ██████╗ ███████╗
██╔════╝██╔══██╗██╔═══██╗██║    ██║██║    ██║██╔══██╗    ██║     ██╔══██╗██╔══██╗██╔════╝
█████╗  ██████╔╝██║   ██║██║ █╗ ██║██║ █╗ ██║███████║    ██║     ███████║██████╔╝███████╗
██╔══╝  ██╔══██╗██║   ██║██║███╗██║██║███╗██║██╔══██║    ██║     ██╔══██║██╔══██╗╚════██║
███████╗██████╔╝╚██████╔╝╚███╔███╔╝╚███╔███╔╝██║  ██║    ███████╗██║  ██║██████╔╝███████║
╚══════╝╚═════╝  ╚═════╝  ╚══╝╚══╝  ╚══╝╚══╝ ╚═╝  ╚═╝    ╚══════╝╚═╝  ╚═╝╚═════╝ ╚══════╝

I build end-to-end systems: native applications, on-device inference, cameras and BLE devices, model-training pipelines, web services, and the infrastructure needed to test and ship them. My strongest work starts with a product or hardware capability and connects every required layer until it works outside a demo.

• On-device and edge AI: quantization, local model runtimes, perception, tracking, and multimodal systems
• Apple platforms: Swift/SwiftUI products, macOS automation, iOS testing, signing, release, and App Store operations
• Wearables and alternative interfaces: smart glasses, BLE, camera/audio relays, silent speech, and sensor-driven interaction
• Agent infrastructure: composable tools, MCP servers, evaluation, orchestration, and machine-to-machine workflows

Larger product systems and internal platform monorepos are mostly private. Older experiments are being consolidated into explicit archives so the public account has a clearer canonical map.

Privacy-focused iOS media processing: metadata removal, watermarking, and fail-closed visual masking.

An AI companion and audio-journaling product focused on personal advocacy, augmentation, and reflection.

A sleep-cycle utility built around 90-minute timing and practical wake-time planning.

Shipping these products has included native implementation, screenshots and localization, signing, CI, TestFlight delivery, App Store Connect automation, and post-release iteration.

Exploring camera-based physiological sensing for phones and wearable cameras.

An earlier real-time self-classification framework and hardware demonstration.

Silent-speech and non-vocal communication research, including visual speech recognition, personalization, and deployment paths for Apple devices.

Reusable service handlers for multiple AI providers, streaming, WebSockets, and provider-specific capabilities.

A small prompt-template registry for predictable, testable LLM interactions.

• Build the thinnest end-to-end system that can be tested in the real environment.
• Keep hardware, model, application, and infrastructure boundaries replaceable.
• Automate repeated work—especially testing, deployment, screenshots, signing, and releases.
• Extract reusable packages after an interface has become real, not merely hypothetical.
• Treat AI coding agents as implementation multipliers; verify the resulting system with tests, device evidence, and operational checks.

Ship end-to-end. Verify on real systems. Extract what becomes reusable.