I built an AI version of myself
- portfolio
- ai-implementation
- customer-success
- build-in-public
For 3.5 years I managed 200+ monthly B2C SaaS accounts. I hit 95% retention and earned two promotions. The work was real. The numbers were real.
But when I started applying for Technical CSM and AI Implementation roles in Israeli high-tech, recruiters needed proof I could build, not only manage. A CV shows you did the work. A portfolio proves you can build what comes next.
So I built one. On my own time, self-taught, using the same AI tools I want to work with professionally.
What is on the site
Three working demos. Each one runs on my own API keys, with rate limiting and prompt injection protection.
- AI Assistant that answers questions about my experience and stack
- Workflow Visualizer that turns a customer journey into a Make.com scenario
- Churn Risk Predictor that scores an account from CSM notes
These are not tools I built at a previous job. AI was never part of my team's stack. I learned to build on my own, driven by curiosity and a hunch that Customer Success and AI belong together.
The stack
- Next.js 15 on Vercel
- Claude API for the assistant and the predictor
- Make.com for automations around the site
- Supabase for auth-gated tools
- Framer Motion and GSAP for the interaction layer
- Claude Code for every line of code
I wrote no code from scratch. I made every decision about what went in.
What shipping this taught me
- Shipping beats polishing. v1 had a Lighthouse score of 62 on mobile. v3 hit 91. The only way to reach v3 was to put v1 live.
- Security is not optional. Rate limiting, server-only keys, input validation, CORS. All three demos have these.
- Self-taught beats waiting for permission. Nobody asked me to learn these tools. I started because I wanted to. That turned into a portfolio, two content brands, and a skill set that did not exist on my CV a year ago.
Where this goes next
I am open to Technical CSM, AI Implementation, and Customer Success roles in Israeli high-tech. I also take on freelance AI Ops work when the fit is right.
The demos are live at michellesartorio.com.