Solo product and engineering effort focused on building a portfolio that feels like a coherent software product rather than a static marketing site. The project couples a public RSC-first frontend with a separate Payload CMS backend, then layers AI retrieval, analytics, and terminal-style interaction patterns on top without breaking maintainability.

• > Built the portfolio as a product-grade monorepo with separate front and back apps instead of a tightly coupled content site.
• > Implemented an RSC-first Next.js frontend with feature-based architecture and terminal-inspired interaction patterns.
• > Added AI-assisted copilot flows with retrieval-ready content synchronization, server-issued visitor identity, and quota hardening.
• > Reduced unnecessary client payload by moving global search data to an on-demand route instead of hydrating the full corpus on every page.
• > Locked anonymous CMS access to published content and moved visit increments behind trusted internal request signing.
• > Wired self-hosted Umami analytics, archive pagination/search, and reusable explorer/tabs primitives into a cohesive user experience.