Western Washington University

• Architected the frontend for SPIE's Digital Library redesign as the founding React/TypeScript engineer on the team, defining the tech stack, module structure, and build standards that moved a large legacy Razor/jQuery codebase to a modern React Islands architecture.
• Designed and built the platform's React/TypeScript component library and owned its highest-traffic pages (article detail and account), authoring a live component showcase that onboarded the rest of the team to React and TypeScript.
• Built the cross-site OAuth handshake and end-to-end session handoff (sign-in, signing, and Redux state) shared across SPIE's web properties, backed by a new Redis caching pipeline.
• Cut page load times by 90% and build times by 50% through Webpack chunking, lazy loading, and pruning redundant assets from the build, and implemented hot module reloading the team had previously been unable to achieve with their SSR React Islands setup.
• Refactored legacy model-bound views into well-structured API and Entity Controllers in C#/ASP.NET MVC.

• Designed and built a novel adversarial framework in Python that outperformed state-of-the-art detection methods, validating results against real municipal network data with cross-dataset validation.
• Developed a novel encoding scheme, rethinking adversarial DNS Exfiltration by using combinatorial mathematics to optimize data transmission through structural patterns rather than payload manipulation.
• Implemented reinforcement learning and LSTM models to optimize system configurations and mimic realistic network behavior, maximizing throughput while evading detection.
• Engineered concurrent data pipelines for simulating, parsing, and validating large volumes of DNS records.

• Evaluated 6+ neural architectures for terrain compression, achieving 30x compression over raw elevation data.
• Built a real-time Vulkan compute shader terrain renderer in Zig with a 7-level clipmap LOD system, hitting 20,000+ FPS on desktop and 300+ FPS on a Raspberry Pi 5. Compression error is within source data noise.
• Designed a training pipeline processing Copernicus GLO-30 GeoTIFF tiles with per-channel INT8 quantization, producing models 50-75x smaller than PNG baselines for global terrain coverage.

• Designed and developed a full-stack course scheduling application serving up to 1,000 monthly users, using bitmask-based conflict detection with backtracking to generate optimized schedules based on timing, compactness, and GPA.
• Built a Go backend with SQLite persistence, a background job scheduler for automated data synchronization, and a concurrent web scraper pulling from WWU's Banner API to maintain up-to-date course catalogs.
• Developed a responsive React/TypeScript frontend with persistent client-side state, real-time course search with relevance scoring, and an interactive calendar interface for browsing and comparing generated schedules.