- Problem
- A static portfolio site can show work, but it can't answer specific questions about it, connect related pieces, or reprioritize itself based on what a visitor actually cares about.
- Her role
- Architecture + build
- Measured
- Source-grounded answers
- Per-call cost tracing
- 30/30 answer evals, with and without an LLM
Hi, I'm Alex OS.Built by Alexandra NacSenior Full-Stack & Applied AI Developer
Ask me about her work — every answer comes with its sources.
Curiosity has a voice.
Ready when you are.
Voice runs on Google Gemini; your audio is processed by Google. Alex OS keeps an anonymous log of conversations (questions, voice transcripts, answers and clicks) to run the assistant. No IP address or device details are stored with it.
Or tell me who you are
Your turn.
Answers come only from published work, with sources. When I don't know, I say so.
Who is Alexandra Nac?
Alex OS
Short version first. Every number below links to how it was measured.
Alexandra Nac is a senior full-stack developer and applied AI developer. She builds AI workflows, retrieval-augmented (RAG) systems and agentic applications, and the web platforms they run on.
What she has shipped
Alex OS
Start with these. I can walk you through the problem, the decisions and what she'd change.
CMS & Web
Alex CMS
Fully dynamic content model with zero hardcoded types.
Next.js · TypeScript · MySQL · mysql2
CMS & Web
Theme DNA Design System
One place to change the brand; 23 templates follow instantly, including AI-generated content.
Next.js · Tailwind CSS · MySQL
How we know it works
Alex OS
Don't take my word for it. Here is what was measured, before and after.
Projects returned on a browse query
18 of 18
- Before
- Typing "projects" into Ask Alex OS returned 1 semantically-nearest project.
- After
- Browse intent is detected; the named type is widened and boosted so the full published set comes back.
verifiedHow it was measuredContext assembly time
1.8s (from 10.5s)
- Before
- Every Alex OS request rebuilt all content blocks from the CMS: ~10.5s.
- After
- Read-through block cache serves assembled blocks: ~1.8s.
verifiedHow it was measuredLLM calls traced
100% (407 of 407)
- Before
- AI generation ran untracked: no record of which model was called, what it cost, or how long it took.
- After
- Each call writes a trace (timestamp, task, provider, model, tokens, cache hits, duration, cost) to an append-only log surfaced on the analytics dashboard — e.g. 407 calls / 299,742 tokens / $1.24 in one development window.
verifiedHow it was measuredHardcoded homepage types
0
- Before
- Homepage sections were hardcoded: adding a content type meant editing React components and redeploying.
- After
- Sections are built from the content-type registry at render time. A brand-new type shows up on the homepage the moment it is published — tested end-to-end with a throwaway type, then removed.
Where she has worked
Alex OS
The roles behind the work, most recent first.
Senior full-stack developer — applied AI and web platforms
One company since 2018, and the work moved with the stack. Web platforms and email infrastructure first; applied AI now: workflow assistants built as prototypes with a human sign-off, recommendation systems, and the evaluation that tells you whether either actually works.
TypeScript · JavaScript · React · Next.js · Node.js · REST APIs
How she thinks and teaches
Alex OS
What she writes about when she isn't shipping, and the workshops she runs.
- Lesson · 2 min read · 13 Claude Skills I pulled out of building this siteEach one started as something that broke here: a fake crawler, an unused vector index, an API key that ran dry. The checklists I wrote while fixing them are now free downloads.
- Lesson · 1 min read · Three bugs that shipped in AI-assisted code, and the gates that catch them now
- Lesson · 1 min read · Twenty-two buttons is the same as no guidance
- Decision · 1 min read · What I let Claude Code do, and what I don't
Alex OS
I can do more than answer. Tap any line and I'll do it.
What Alex OS can do
Say it in a voice call, type it below, or use what's on the page. Tap a typed example to try it. Tap any line to try it.
Say or type
Ask about her work
Anything about Alexandra's projects, skills, career or writing. Answers come only from what's published here, with sources.
Say it
Quiz me
A voice quiz: Alex OS asks your name, then four questions about Alexandra's work, tells you after each one whether you were right and why, and shows your scorecard.
Say it
Show me something
Opens a door here, or takes you to a page. By voice, going to another page ends the call.
Say it
A quick guided tour
Alex OS walks you through two to four parts of the site that fit you, asking before each step.
Say or type
Compare two things
Side by side: what each one is, how they differ, when each makes sense.
Say or type
Send Alexandra a message
Write or say it; Alex OS reads it back and only sends it when you confirm. Leave an email if you'd like a reply.
Questions people ask me
Alex OS
The ones I hear most, answered from her published work.
Who is Alexandra Nac?
Alexandra Nac (Alex) is a senior full-stack developer working in applied AI. She builds with AI and builds AI into products: assistants, retrieval and recommendation systems that sit inside real workflows, plus the web platforms they run on. This site is her working example, and Alex OS is the assistant that answers questions about her.
How does Alexandra work with AI?
As a pair, not a code generator. She sets the architecture, writes the scope, lets the model implement in small steps, reviews every change, and keeps tests and evaluation suites as the gate. The systems on this site were built that way, and the full loop is written up on the How I work with AI page.
What does Alexandra Nac work on?
Four things: applied AI (assistants, RAG and retrieval, agentic workflows, evaluation and cost tracing), full-stack systems (frontend, backend, APIs, CMS architecture, data workflows), UI/UX thinking (interfaces that guide people and reduce confusion), and the working tools around her own delivery, such as evaluation suites and review workflows.
What kind of projects has Alexandra Nac shipped?
What she has built: an email helpdesk responder, an insurance policy renewal assistant, a product recommendation engine with the SDK that measures it, email delivery infrastructure and WordPress platforms. Plus the systems behind this site: Alex OS, Alex CMS and Theme DNA, each published with its decisions and trade-offs.
What is Alex OS?
Alex OS is Alexandra's assistant and the reasoning layer behind this site. It answers questions about her from her published content, with sources, and runs as a separate service between the CMS and the public pages. It is enhancement-only: if it goes offline, the site still works.
How can I contact Alexandra Nac?
Through the contact form on this site, by email, or via the GitHub and LinkedIn links in the footer. If you'd rather ask something first, Alex OS answers from her published work.
Alex OS
And the person who built me.
Who built Alex OS
Alex OS is her retrieval-augmented assistant: it answers only from her published work, cites the sources behind every answer, and can open pages, send messages and talk by voice.
Selected work
- Alex OSA FastAPI reasoning and retrieval layer that sits between the CMS and the public site — grounded answers, journey-aware recommendations, and a relationship-graph reasoning pass, not an open-ended chatbot.↗
- Alex CMSA Next.js CMS with a fully dynamic content-type registry — no hardcoded post models — plus a journey-based homepage, an orchestrator workflow, and per-item content scoring.↗
- Theme DNA Design SystemA database-driven design token system ("Theme DNA") that every public template consumes — change the brand once, and 23 templates restyle themselves. No hardcoded colors anywhere.↗
Expertise
- Applied AI Systems
- Dynamic CMS Architecture
- RAG and Knowledge Retrieval
- AI Workflow Automation
- API and Backend Integration
- AI Evaluation and Content Intelligence
- Full-Stack Product Development
- AI Observability & Cost Governance

