Native iOS experiences, autonomous agent workflows, device bridges, and verification-first automation. My current focus is CARINA and MAYA: a family of tools designed around privacy, human control, and reliable execution.

Two native iOS keyboard experiences in one verified suite.

• CARINA uses Apple's on-device Foundation Models for user-requested replies and rewrites on supported iOS 26+ devices.
• MAYA delivers fast deterministic, context-aware prepared replies.
• Both keyboard extensions keep Full Access disabled and leave final insertion to the user.
• Includes dual-project CI, privacy checks, evidence-based documentation, and four specialized repository agents.

A control surface for agent operations, system visibility, and practical AI workflows.

An orchestration layer for routing work, coordinating tools, and building repeatable agent behavior.

A macOS-native accessibility aid that turns required clicks and approvals into large, high-contrast, spoken guidance.

Observe the real system -> build the smallest complete change -> verify it -> document the evidence -> repeat

• Real integrations over simulated behavior
• Privacy boundaries stated precisely
• Native Swift and macOS tools where they make the system simpler
• Explicit error handling, logging, and reproducible checks
• Human approval for consequential actions

Swift SwiftUI UIKit Foundation Models Python GitHub Actions macOS automation iOS extensions API integrations

My flagship repositories include concise project context, automated verification, and focused agent roles. In CARINA x MAYA TYPE, the agent team covers builds, privacy review, documentation evidence, and release preparation while keeping publication and App Store actions under human control.

Open an issue in the relevant repository with the problem, expected outcome, and environment. I am especially interested in privacy-first AI, accessibility, iOS tooling, and dependable agent systems.