AI Architect for Regulatory Document Processor

We are building Regulatory Watchtower – a private AI system designed to monitor, analyze, and interpret continuously changing regulatory documents (e.g. pharma, compliance, legal texts).

The system is not a demo. It is meant to be:
explainable
auditable
resistant to hallucinations
suitable for regulatory environments

This is a real production-grade AI system, not a chatbot toy.

Your Role
You will act as an AI / LLM Architect, responsible for:
designing the AI architecture
implementing a robust RAG pipeline
ensuring traceability, source integrity, and answer reliability
supporting the transition from POC to production
You will work directly with the product owner (business + regulatory background).

Required Skills
Strong Python (production-level)
Hands-on experience with LLMs:
OpenAI / Azure OpenAI and/or local models (LLaMA, Mistral)

RAG architectures:
embeddings, chunking strategies, retrieval tuning
vector databases (FAISS, Qdrant, Weaviate, Pinecone)

Document processing:
PDFs, OCR, regulatory texts
Prompt engineering + response evaluation
FastAPI (or similar)
Git, Docker

Nice to Have
Experience with regulatory, legal, pharma, compliance, or financial documents
AI Agents / workflow orchestration
Private / on-prem AI setups
Hallucination mitigation strategies
Monitoring & evaluation of LLM outputs

What We Value Most
System thinking over hype
Ability to explain why an answer is generated
Awareness of LLM limitations and risks
Clean documentation and architectural reasoning
If your instinct is “this needs guardrails” – you’re our person.

Engagement Model
Start with a 4–6 week POC
Extend into iterative development
Flexible workload
Competitive freelance rates (senior level)