AI / ML Engineer (RAG + Extraction)

About the role

Purple8 ships three built-in RAG modes (flat vector, graph-augmented, hybrid) with autonomous corpus profiling and tuning, plus DocIntel — a document-intelligence service that turns 70+ formats into queryable knowledge via a two-pass extraction pipeline. You'll own retrieval quality, extraction accuracy, and the AI hooks in the workflow engine.

What you’ll own

• Own the hybrid RAG retrieval path and its self-tuning/profiling.
• Improve entity + relationship extraction across domain profiles.
• Build evaluation harnesses that measure retrieval and extraction quality objectively.
• Integrate LLM providers behind a clean, optional-by-default abstraction.

What we’re looking for

• Hands-on RAG engineering — chunking, embeddings, reranking, fusion.
• Practical NER / information-extraction experience.
• Rigor about evaluation: F1, MRR, token/recall trade-offs.

Nice to have

• GLiNER / transformer fine-tuning (LoRA adapters).
• Experience with agentic / MCP tool interfaces.