Senior Full-Stack Developer (AI Document Pipelines & AWS) - WW - FT

Role Overview

We are seeking a seasoned, production-proven Senior Full-Stack Developer to design and deliver Phase 1 of an enterprise AI-powered quoting and document extraction tool for a leading Canadian manufacturer.

This role is NOT about building chatbots, basic RAG search, or Q&A interfaces. You will architect end-to-end automated pipelines that turn complex, messy, unstructured customer documents (PDFs, multi-line RFQs, email attachments) into strictly validated JSON schemas and structured data for enterprise system consumption (ERPs, SAP, custom CRMs).

Note: Candidates based in South America (LATAM) with strong English communication and aligned overlap are encouraged to apply.

Key Responsibilities

• AI Extraction & Evals Pipeline: Architect AWS serverless workflows (Lambda, Aurora DB, API Gateway) that ingest messy PDFs/documents, leverage Anthropic Claude on Amazon Bedrock for structured data extraction, and enforce strict schema validation, retry logic, and fallback routines.
• Automated Evals & Quality Control: Design and implement automated evaluation (eval) frameworks to measure and benchmark non-deterministic LLM output accuracy, schema drift, and hallucination rates, alongside standard unit/integration test suites.
• Angular Web App & Document Generation: Build a responsive Angular (TypeScript) web application enabling sales reps to review, edit, and manage extracted line items, linked to a Node.js engine for programmatic Word (.docx) quote generation and Outlook export workflows.
• (Bonus Track / Secondary Persona): Architect custom Model Context Protocol (MCP) servers from scratch and/or integrate AWS Bedrock AgentCore MCP Gateway (Cognito, DynamoDB, Lambdas).

Must-Have Requirements (Hard Criteria)

1. Shipped AI Products in Production: Proven history of building and shipping production software powered by LLMs or foundation models. (Note: Using AI coding assistants like Claude Code, Cursor, or Copilot does NOT fulfill this requirement).
2. Document Extraction over Chatbots: Concrete experience building LLM pipelines that parse unstructured documents (PDFs, invoices, RFQs) into rigid, structured outputs (JSON/APIs) for ERP/CRM consumption.
3. Provable Dual-Testing Rigor:
   • Automated unit/integration tests for deterministic business logic.
   • Automated evaluation (eval) suites for non-deterministic AI outputs (e.g., accuracy, schema compliance, fallback validation).
4. Production Stack (Proven in Work History): Hands-on engineering depth in:
   • Frontend: Modern Angular (TypeScript), state management, complex enterprise UI forms.
   • Backend: Node.js / TypeScript.
   • Cloud & Serverless: AWS Lambda, API Gateway, Amazon Aurora (PostgreSQL/MySQL).
5. Document Processing: Hands-on experience with PDF parsing libraries and programmatic Word generation (docxtemplater, docx, etc.).

Nice-to-Have / Specialized Skills

• Hands-on development of custom MCP (Model Context Protocol) servers built from scratch.
• Experience with AWS Bedrock AgentCore MCP Gateway (Cognito, DynamoDB, Lambda integrations).
• Integration with enterprise ERP APIs (SAP, Salesforce, Dynamics).