AI Engineer — RAG Pipeline, Voice Agent, Python, n8n, Supabase

Compensation & Growth Path:
Phase 1 — AI Engineer (Project-based) $1,000 USD fixed for the initial project — paid via Deel milestone escrow
Phase 2 — Senior AI Engineer (Full-time) $1,200-1,500/month for the right candidate begins after successful project delivery
Phase 3 — Lead AI Engineer (Leadership) $2,500-3,000/month for the right person who grows into managing and scaling our developer team
Compensation scales as the business grows. We take care of the people who take care of us. If you want to build a career not just complete projects this is the opportunity.

About Us: Empire42 is a US-based AI agency that builds custom AI clones for health and fitness influencers. We extract a coach’s entire content library, build a structured knowledge base from it, and deploy a voice-enabled conversational AI agent that responds in the coach’s voice so their high-ticket clients get instant answers between coaching sessions without the coach lifting a finger.
We’re building something real and we want someone who wants to grow with it.

About the Project: Your first project is building a complete conversational AI voice agent and RAG pipeline for a fitness influencer with two-plus years of YouTube content. The agent needs to be voice queryable with low latency audio responses clients speak to it naturally and get responses in the coach’s voice within seconds.

What You’ll Build:
• YouTube transcript ingestion pipeline using youtube-transcript-api and Whisper API
• Vector knowledge base in Supabase pgvector with HNSW indexing
• Python-based chunking and embedding pipeline using OpenAI text-embedding-3-small or Voyage AI
• RAG retrieval pipeline with cosine similarity dense search first, hybrid BM25 sparse search ready for V2
• Claude API integration with custom system prompt engineering for voice and personality matching
• Mem0 memory layer for cross-session client continuity and long term context
• Real-time voice pipeline Deepgram STT for speech to text, ElevenLabs streaming TTS for low latency audio responses
• n8n workflow orchestration connecting all pipeline components
• Branded web app frontend voice recording interface, audio playback, user authentication, mobile responsive generated via Claude Code and integrated into the pipeline
• End to end voice latency optimization to under 3 seconds from input to audio response

Required Skills:
• Python — strong proficiency, clean production-ready code
• RAG pipeline architecture chunking strategy, embeddings, vector retrieval, context injection
• Vector databases Supabase pgvector or Pinecone, hands-on production experience
• Claude API or OpenAI API — direct API integration experience, not just wrapper tools
• Voice pipeline — Deepgram STT, Whisper API, ElevenLabs streaming TTS
• n8n — workflow automation, API chaining, webhook handling
• Mem0 or equivalent memory layer — cross-session context management
• REST API integration — webhooks, async handling, error management
• LangChain or LlamaIndex — production pipeline experience preferred
• Supabase — database setup, pgvector configuration, query optimization
• Claude Code or AI-assisted development — proficiency strongly preferred
• Web app deployment — hosting, environment configuration, domain setup

Nice to Have:
• BM25 hybrid search implementation
• Langfuse or LangSmith RAG monitoring and tracing
• Twilio WhatsApp API integration
• Conversational AI agent development for real clients
• Voice cloning pipeline experience
• ElevenLabs Professional Voice Cloning
• React or Next.js frontend experience

Education: Bachelor’s degree in Computer Science, Software Engineering, or related field required. Master’s degree preferred.

Work Style: This is a fully async remote role. You set your own hours and work when you’re most productive. No night shifts, no US timezone requirements. Responsive within 4-6 hours during your chosen working hours. We don’t require US timezone availability but we do expect timely communication and proactive updates on progress. Weekly check-in call at a mutually agreed time. We care about output and delivery — not when you’re online.

Growth Path: For the right person this role has a clear and well-compensated advancement path:
• Phase 1 — Project-based AI Engineer. Prove your skills on the first build. $1,000 fixed.
• Phase 2 — Full-time Senior AI Engineer. Own all client builds at $1,200-1,500/month.
• Phase 3 — Lead AI Engineer. Hire, onboard, and manage our growing developer team as we scale to multiple clients per month. $2,500-3,000/month.
Compensation scales as the business grows. We take care of the people who take care of us. If you want to build a career not just complete projects — this is the opportunity.

What We’re Looking For:
Someone who has actually built and shipped RAG pipelines and conversational AI voice agents in production — not just completed courses. We need someone who can take a technical spec and execute independently without hand-holding.
Production experience is required. Please do not apply if you have only completed courses or personal projects.
Please include:
• Links to GitHub repos or live production projects
• A short Loom video (5-10 minutes) walking through a relevant RAG pipeline or voice agent you built
• Brief description of your specific experience with real-time voice pipelines
• Answer this question in your application: “Describe how you would handle chunk overlap in a RAG pipeline and why it matters for retrieval quality”
Applications without a Loom video, GitHub portfolio, and answer to the technical question will not be considered.

Project Scope and Timeline:
• Week 1 — Content ingestion pipeline and knowledge base live
• Week 2 — RAG retrieval and Claude API clone core complete
• Week 3 — Voice pipeline integrated and latency optimized
• Week 4 — Web app frontend deployed, revisions complete, handoff
• You must be able to dedicate 20+ hours per week to this project

Milestone Payment Structure via Deel Escrow:
• $400 released — Knowledge base live and retrieving correctly
• $400 released — Voice pipeline end to end working under 3 seconds
• $200 released — Web app deployed and final delivery approved by Empire42
All payments processed through Deel milestone-based escrow. Please only apply if you are comfortable with this payment structure.

What Success Looks Like:
• RAG pipeline retrieves accurate relevant chunks for any client query
• Conversational AI voice agent responds in the coach’s voice, tone, and frameworks
• End to end voice latency consistently under 3 seconds
• Mem0 memory layer correctly recalls client context across sessions
• Escalation logic triggers correctly on sensitive or off-limits topics
• Branded web app is live, mobile responsive, and client-ready
• Coach personally approves the clone as an accurate representation of their brand

To Apply:
Send your application with:
1 Loom video walkthrough of a RAG pipeline or voice agent you built in production
2 GitHub or portfolio links showing relevant projects
3 Brief paragraph on your real-time voice pipeline experience — Deepgram, Whisper, ElevenLabs
4 Your weekly availability in hours
5 Answer to the technical screening question above