AI/ML Engineer

• Build and tune the RAG / grounding layer over a PostgreSQL-backed knowledge store with row-level security.
• Implement generators that compose retrieved evidence into a fixed output contract (structured slide and document formats).
• Implement a fail-closed check engine — named, individually reportable pass/fail tests that block delivery of ungrounded output.
• Integrate a real-time evidence service drawing on external public and licensed sources, with dated, source-linked attribution.
• Integrate with an enterprise-governed model gateway rather than calling providers directly; work within token budgets.
• Build artifact export to PPTX / DOCX / PDF, carrying workflow-contract metadata through to the rendered file.
• Contribute to the current-state assessment of an existing codebase and its test suite.

REQUIRED SKILLS AND EXPERIENCE

• 6+ years software engineering with 2+ years shipping LLM-backed features to production — not prototypes.
• Strong Python; production API development (FastAPI or equivalent).
• Demonstrated end-to-end RAG work: chunking and retrieval strategy, embeddings, reranking, citation integrity, and measuring retrieval quality.
• Evaluation discipline — you have used an eval harness in anger and can discuss hallucination rate, retrieval precision, and cost/latency as tracked numbers.
• LLM orchestration frameworks, and programmatic document generation (python-pptx, python-docx, or equivalent).
• Containerized deployment and CI/CD.

PREFERRED

• Multi-agent / persona-based orchestration.
• Working through an enterprise model gateway or proxy under governance and token budgets.
• PostgreSQL row-level security or other entitlement-aware data access.
• Semantic caching and cost-per-query optimization.