Most people bolt an AI chatbot onto their site and call it a day. That's not what I built.
The Odin interface on this site isn't a generic wrapper around an API. It's trained on my actual work — my methodologies, my project history, my approach to automation and finance operations. Ask it about measurable results. Ask it about process design. It knows.
Here's what I learned building it in a weekend:
1. The knowledge base is everything. The system prompt isn't hand-wavy personality fluff. It's structured data about real deliverables, real metrics, real tools I've used. Every claim the AI makes traces back to something concrete.
2. Rate limiting matters. The chat endpoint has per-IP throttling. Not because I expect abuse — because I want the experience to stay fast and reliable even if a thread gets popular.
3. Design sets the expectation. The navy/maroon/cream palette, the glass morphism, the grain texture — these aren't decorative. They signal "this is a professional who cares about craft." Before the AI says a word, the visitor already trusts the experience.
The site is live now. Try the chat. Ask about automation wins, compliance work, or the tools I use. The answers will surprise you — not because the AI is clever, but because the person behind it did the work.
This is the first post on the blog. More coming as I build in public.