This Portfolio — Built Entirely by Agentic AI
A full-stack, AI-powered portfolio and lead-generation platform — this very site — built end-to-end by an autonomous AI coding agent (Claude Code) rather than hand-written line by line. Includes a conversational AI assistant, a complete admin CMS, and live third-party integrations, all implemented, tested, and shipped through natural-language direction.

Screenshots


Built With Agentic AI
This site itself was built end-to-end by an autonomous AI coding agent directing implementation, not hand-written line by line. Here's roughly how that compares to a traditional build.
Traditional Development
- Time
- ~90 hours
- Est. Cost
- ~$4,500 (at $50/hr)
Agentic AI (This Build)
- Time
- ~10 hours
- Tokens Used
- ~8.5M tokens
- Est. Cost
- ~$650
The Goal
Prove that an agentic AI coding assistant can take a project from a written specification to a fully working, production-deployed product — architecture, UI, backend, database, third-party integrations, and admin tooling — with a human directing scope and reviewing outcomes rather than writing implementation code by hand.
Challenges Faced
- Keeping a large, evolving codebase consistent across dozens of features added incrementally in natural language, without regressions
- Making every external integration (AI provider, email, image storage, database) degrade gracefully when unconfigured, so the app never breaks for missing credentials
- Diagnosing real environment issues — Windows file locks during Prisma regeneration, slow-filesystem dev-server compile times, a genuine hydration bug — with the same tools available to a human engineer
- Migrating the database from local SQLite to a production Postgres instance mid-project without losing schema integrity
- Balancing AI response quality against Gemini API token and quota costs in the live chat assistant
Solutions
- Verified every feature end-to-end in a real browser — not just type-checks — before calling it done, catching real bugs like a hydration mismatch and an ambiguous test locator before they shipped
Key Features
- Conversational AI assistant with RAG-lite knowledge retrieval, project-discovery flow, and interactive mockup generation
- Password-protected admin CMS for projects, services, profile, and AI knowledge — with live sync into what the assistant knows
- AI-assisted bulk knowledge import and automatic logging of questions the assistant couldn't answer
- Multi-key Gemini failover with graceful degradation to a deterministic rule-based provider
- Amazon S3 image uploads for project screenshots with manual-URL fallback
- Lead capture with AI-generated project summaries and email notifications
Technologies
Have a similar project in mind?
Tell my AI assistant what you're looking to build.
Start a Conversation