Experienced Ai/Backend Engineer for Saas Proposal Generation Module

We are seeking an experienced AI/Backend Engineer to develop a production-ready AI proposal generation module for our SaaS platform. This is a critical role for a serious, long-term collaborator, not a quick fix.

This project is focused on building a structured AI pipeline integrated into a Node.js backend, specifically not a chatbot project.

Objective:
The primary goal is to enable users to upload two types of documents:
1. A construction project PDF (typically 10-30 pages, featuring mixed layouts, tables, and notes).
2. A pricing table (provided in Excel/CSV format with standardized items and units).

The system must perform the following functions:
*  Parse the uploaded pdf documents, including ocr, layout analysis, and table extraction.
*  Extract structured data such as quantities, scope items, and relevant notes using a Large Language Model (llm) with strict json schema enforcement.
*  Deterministically map the extracted items from the PDF to the standardized pricing table.
*  Calculate all totals accurately outside of the LLM to prevent any hallucinated pricing.
*  Generate a structured proposal document, converting it from docx to pdf format.
*  Return the final generated PDF file for user download.

Technical Requirements:
*  The backend must be developed using Node.js and Express.
*  Proven experience with OpenAI APIs, specifically for structured outputs, function calling, and JSON schema enforcement.
*  Proficiency with document parsing tools such as Google Document AI, LlamaParse, Azure Form Recognizer, or similar solutions.
*  Implementation of a robust validation layer to prevent hallucinated or unmapped pricing rows.
*  Adherence to a clean, modular service architecture.
*  Comprehensive error handling and logging mechanisms.
*  Design for scalable asynchronous processing.

Deliverables:
*  A complete, fully functional backend module for AI proposal generation.
*  Clear and comprehensive API documentation, including request and response examples.
*  Integration of a proposal template.
*  Detailed deployment instructions.
*  An estimation of AI processing costs per document.

Initial volume is expected to be approximately one proposal per day. When applying, candidates should detail their chosen parsing tool and rationale, methods for enforcing strict structured extraction, strategies to prevent hallucinated pricing errors, estimated hours and milestones, and examples of similar work.