AI Engineer - RAG Pipeline, Claude API & Prompt Engineering -Hospitality SaaS MVP - Contract to Hire
We are building a B2B SaaS platform for the hospitality industry from scratch. The AI layer is the core of the entire product — it is what hotels pay for. We need an experienced AI engineer who has built real production RAG systems, not just tutorial projects. Full technical brief shared after NDA is signed.
What you will build:
The complete AI layer of the platform including a RAG pipeline using pgvector on Supabase, a dynamic multi-layer system prompt architecture, an automated knowledge base builder that scrapes hotel websites and extracts Q&A pairs using Claude, confidence scoring and human escalation logic, multilingual AI responses across 6 languages, and an AI-powered campaign message generator. Full spec provided after NDA.
AI Stack — already decided, non-negotiable:
LLM: Anthropic Claude API (claude-sonnet) — not OpenAI, not Gemini
Embeddings: OpenAI text-embedding-3-small or Voyage AI
Vector Search: pgvector via Supabase
Web Scraping: Firecrawl API or Cheerio
Queue Management: BullMQ or similar
Language: Python or Node.js
You must have:
Production RAG pipeline experience — you have built and shipped one to real users
Strong prompt engineering portfolio with measurable outcomes
Hands-on experience with vector databases
Experience with the Claude API specifically — not just OpenAI
Clear understanding of hallucination mitigation in customer-facing AI products
Multilingual LLM testing and quality tuning experience
EU timezone availability
You must answer both screening questions to be considered. Applications without answers will not be reviewed.
Engagement details:
Duration: 12 weeks
Budget: Fixed price, milestone based
Milestone 1 — 30% at week 2 on RAG pipeline working end to end
Milestone 2 — 30% at week 6 on knowledge base builder live and multilingual confirmed
Milestone 3 — 40% on final delivery meeting all acceptance criteria
Start date: Immediate
Full brief and NDA shared with shortlisted candidates only
We are serious, organised, and ready to start. We expect the same from you. No agencies. Individual freelancers only.
Apply tot his job
Apply To this Job
What you will build:
The complete AI layer of the platform including a RAG pipeline using pgvector on Supabase, a dynamic multi-layer system prompt architecture, an automated knowledge base builder that scrapes hotel websites and extracts Q&A pairs using Claude, confidence scoring and human escalation logic, multilingual AI responses across 6 languages, and an AI-powered campaign message generator. Full spec provided after NDA.
AI Stack — already decided, non-negotiable:
LLM: Anthropic Claude API (claude-sonnet) — not OpenAI, not Gemini
Embeddings: OpenAI text-embedding-3-small or Voyage AI
Vector Search: pgvector via Supabase
Web Scraping: Firecrawl API or Cheerio
Queue Management: BullMQ or similar
Language: Python or Node.js
You must have:
Production RAG pipeline experience — you have built and shipped one to real users
Strong prompt engineering portfolio with measurable outcomes
Hands-on experience with vector databases
Experience with the Claude API specifically — not just OpenAI
Clear understanding of hallucination mitigation in customer-facing AI products
Multilingual LLM testing and quality tuning experience
EU timezone availability
You must answer both screening questions to be considered. Applications without answers will not be reviewed.
Engagement details:
Duration: 12 weeks
Budget: Fixed price, milestone based
Milestone 1 — 30% at week 2 on RAG pipeline working end to end
Milestone 2 — 30% at week 6 on knowledge base builder live and multilingual confirmed
Milestone 3 — 40% on final delivery meeting all acceptance criteria
Start date: Immediate
Full brief and NDA shared with shortlisted candidates only
We are serious, organised, and ready to start. We expect the same from you. No agencies. Individual freelancers only.
Apply tot his job
Apply To this Job