AI Agent Evaluation Developer for RAG

# AI Agent Evaluation Developer for RAG Document Workflow POC We are looking for a developer to build a small document agent and evaluate it using Braintrust or Plumloom. We will provide one large PDF or Markdown document. The agent should: * Answer questions about the document * Search the document multiple times when needed * Refine searches when evidence is incomplete * Generate a structured report or brief with citations * Capture the full agent trace Google ADK is preferred, but LangChain or another open-source framework is acceptable. You are free to design the chat interface using AI tools of your choice. ## Evaluation Work * Create 15 to 20 test cases * Define expected answers, report sections, or behavior * Evaluate correctness, grounding, completeness, citations, and search trajectory * Review failures and identify their cause * Turn important failures into regression tests * Re-run evaluations after changes * Show one before-and-after improvement * Document what the evaluation platform measures well and where it needs improvement ## Deliverables * Complete source code * Setup instructions * Working document workflow * Exported traces * Evaluation dataset and results * Short findings report *AI coding assistants are allowed. The developer must use their own model credentials.* Duration: 1 week --- *This will be a milestone-based project managed and paid through Upwork.*