From idea → working AI product
How This Site Was Built
This site is an AI case study.
I wanted more than a digital resume. I wanted to build an AI-powered experience that lets hiring leaders explore my background, evaluate role fit, and see how I approach AI product development in practice.
Built and run with
From idea to production
The build, step by step
Define the Experience
What should a hiring leader actually be able to do?
Architect with AI
Translate the idea into product requirements, data structure, UX, and AI architecture.
Prototype Rapidly
Explore interface and interaction ideas before overbuilding.
Build & Iterate
Use AI coding tools to implement, debug, and refine quickly.
Ground the Intelligence
Connect AI answers to verified career evidence instead of allowing generic or invented claims.
Run AI Locally
Use Qwen + Ollama for controlled, cost-conscious public inference.
Harden for Real Users
Continuously test grounding and latency; design and load-test caching, traffic protection, and retry logic ahead of rollout.
Ship & Learn
Deploy, observe usage, gather feedback, and keep improving.
Product principles
The decisions behind the build
Grounding over cleverness
The AI should say what Julie has actually done — not what merely sounds plausible.
Local first
Use local inference where practical to control cost and experiment with the architecture.
Design for the wait
Instead of hiding AI latency, make the experience communicate useful progress while the model works.
Build for reality
Caching, rate limits, traffic protection, and graceful failure matter when an experiment becomes a public product.
AI accelerated the build. It didn’t replace product judgment.
Division of labor
What Julie did — and what AI accelerated
Julie
- Product concept
- Experience design
- Architecture decisions
- Career-data structure
- Grounding requirements
- Quality standards
- Testing strategy
- Technical/product tradeoffs
- Deployment decisions
AI Collaborators
- Rapid prototyping
- Code generation
- Implementation acceleration
- Debugging support
- Test support
- Design iteration
Takeaways
What this build reinforced
The model isn’t the product.
The surrounding system — data, grounding, UX, controls, and feedback loops — is what makes AI useful.
Speed changes the operating model.
When AI lets you change a product daily, testing and release practices have to evolve too.
The architecture should absorb change.
Models and tools will change. The product shouldn’t have to be reinvented every time they do.