Stellenbeschreibung:

About Foodamigos

We help restaurants get more orders and reservations with a full website builder and editor. Today our system is template/section-based (no AI). We want to add an AI “type-and-get-a-site” flow so restaurant owners can generate slick, on-brand sites from a prompt—then contact us to publish.


Project Overview (What we want)

Build a Lovable-style experience on top of our existing builder:
• User types a prompt (e.g., “Modern salad bar, like Sweetgreen, with online ordering”).
• System interviews for 4–5 details (name, cuisine, vibe, primary CTA, colors).
• Generates a site using our components, our design tokens, and optional “style clone” from a reference site (legally safe: copy style, not assets).
• Shows an instant preview; “Publish” is gated behind a Get-in-Touch form.
We need a consultant who can architect, hands-on prototype, and coach our team to production.


Scope & Responsibilities

Architecture & Plan
• Define end-to-end flow: prompt → tokenized design → component composition → diff/patch → deploy → preview.
• Recommend services (LLM provider, deploy, storage) and security guardrails.

Hands-on Prototype (PoC)
• LLM pipeline: Anthropic Claude or OpenAI (your call) with prompts/guardrails (no inline hex, only our tokens).
• Patch/diff workflow: generate code/JSON diffs and apply safely (Freestyle/“Adorable” style or similar).
• Deploy: push to Git (or our repo) and trigger Vercel/Netlify for a live preview URL.
• Design tokens: extract our current theme (colors, fonts, spacing, radii, motion) into a single source of truth (Tailwind/CSS variables).

“Style-Clone” Microservice (optional but desired)
• Given a reference URL (e.g., Sweetgreen), fetch public HTML/CSS/screenshots (respect robots.txt).
• Distill palette, fonts, spacing/radii into tokens; map to closest Google fonts if needed.
• Feed tokens to the LLM so output uses our components with that “vibe,” not copied assets.

Integrations
• Keep/extend our modules: online ordering & table reservations as pluggable components.
• Basic image/copy helpers (e.g., DALL·E or Stable Diffusion for hero/dish photos, SEO meta + Schema.org).

Knowledge Transfer
• Document prompts, token schema, service boundaries, and deployment.
• Pair with our engineer(s), code review, and handover.


Required Experience


• Strong TypeScript/Node and React/Next.js (or Vue/Svelte with similar patterns).
• Tailwind CSS + design-token systems (CSS variables, theme config).
• Building with LLMs for code/content generation (Anthropic/OpenAI) and prompt guardrails.
• Headless browser tooling (Puppeteer/Playwright) and light HTML/CSS parsing for palette/font extraction.
• Git-based patch/diff workflows; CI/CD to Vercel/Netlify/Cloudflare Pages.
• Practical security: API key handling, rate limiting, safe scraping (robots.txt), and IP transfer.


Nice to Have


• Supabase/Neon/Postgres; Resend/SendGrid; S3/R2 storage.
• Experience cloning design style legally (copy style tokens, not assets).
• Experience with restaurant tech (ordering/reservations) a plus.


Our Current/Target Stack


• Web: React/Next.js + Tailwind (we can adapt if needed).
• Infra: GitHub + Vercel (open to Netlify/CF Pages).
• Data: Postgres/Supabase (or we’ll align with your proposal).
• AI: Claude 3.x or GPT-4o; image gen optional.


Engagement Details


• Engagement: 2–4 weeks, part-time or full-time; consulting + hands-on build.
• Location/Time: Remote; some overlap with Europe/Berlin time.
• Budget: Open to hourly or milestone-based. Please propose.

NOTE / HINWEIS:

EN: Please refer to Fuchsjobs for the source of your application

DE: Bitte erwähne Fuchsjobs, als Quelle Deiner Bewerbung