Product-minded software engineer building AI-native mobile experiences and reliable backend systems.

I care about the parts of AI products that live beyond the demo: trustworthy context handling, structured outputs, deterministic fallbacks, privacy boundaries, testable architecture, and a calm user experience.

Lumi is an iOS assistant that brings calendars, reminders, weather, user-approved memory, and conversational planning into one daily workflow.

I lead the product and engineering work across:

• a native SwiftUI client using EventKit, WeatherKit, Core Location, Speech, and UserNotifications;
• a FastAPI backend with strict Pydantic contracts, Supabase authentication and persistence, and provider abstractions;
• structured calendar/task mutations with explicit review before user data changes;
• deterministic scheduling, fallback, and cache paths that keep the product useful when AI is unavailable;
• automated coverage across iOS domain logic, API contracts, repositories, authentication, and provider failure modes (817 iOS tests and 286 backend tests in the current verified build).

The source repository is private during active product development and is available for interview review.

OpenMP parallelization and strong-scaling analysis of a 2D Jacobi stencil, including reproducible benchmarks, byte-level correctness checks, bandwidth analysis, and an explanation of memory-bound scaling behavior on Apple silicon.

A reproducible SZ3 rate–distortion study on Hurricane ISABEL data with error-bound sweeps, independent PSNR verification, compression metrics, and generated analysis plots.

Swift · SwiftUI · Python · FastAPI · Supabase/PostgreSQL · LLM systems · C/OpenMP

• Start with the user workflow and define what must remain deterministic.
• Keep external providers behind small, replaceable interfaces.
• Validate AI output as untrusted data before it reaches product state.
• Prefer reproducible experiments and measurable results over unsupported claims.
• Use automated tests to protect domain rules and integration boundaries.