AI Engineer - Document Extraction and PII Redaction (Python)

Build the privacy-first AI backbone of a fast-moving financial automation platform. This is a senior individual contributor role with direct ownership over document extraction architecture - no committee approvals, no sprawling org layers.Salary: $150,000 - $200,000 / year depending on experience

Location: Fully Remote - US-based candidates only What you’ll do

• Design and scale high-accuracy document extraction pipelines that process complex PDFs, scans, tables, and financial statements in production
• Build and maintain programmatic PII masking and redaction layers that enforce strict data-privacy boundaries before any data reaches an external LLM API
• Integrate and prompt frontier LLM and vision APIs (OpenAI, Anthropic, Azure OpenAI) for structured financial data extraction
• Enforce downstream reliability with Pydantic models, JSON schema, or native function calling to guarantee schema-validated outputs
• Own rigorous evaluation: build test sets, regression benchmarks, and accuracy metrics that continuously measure and improve pipeline performance
• Drive key architectural decisions on document AI tooling (Azure Document Intelligence, AWS Textract, Google Document AI, or equivalent)
• Partner with platform engineers to harden AI pipeline security and satisfy financial-data compliance requirements

What you need

• Claude Code proficiency (strict requirement): Daily, production-level hands-on experience using Claude Code (Anthropic’s agentic CLI) for development, refactoring, and codebase navigation - candidates without this will not advance
• Expert Python engineering: Production-grade Python with a proven track record building and scaling enterprise AI/ML pipelines
• Enterprise document extraction: 12+ months of recent, hands-on production experience with Azure Document Intelligence, AWS Textract, or Google Document AI
• Frontier LLM implementation: Current production-level expertise with frontier LLM and vision APIs (OpenAI, Anthropic, or Azure OpenAI)
• PII guardrails: Direct experience designing programmatic PII detection and automated redaction within live AI data flows

Nice to have

<li class=”Z1qcYe” style=”border-bottom-color:rgb(10, 10, 10);border-bottom-width:0px;margin-bottom:12px;margin-right:0px;margin-top:0px;text-decoration:none;” data-sfc-cp=”” jsaction=”” jscontroller=”oSLmPe#HokZ0d” data-sfc-root=”ep” jsuid=”AAroQc_3n” data-hvei