DocAI Engineer

Pharma batch records, invoices, forms and engineering drawings still arrive as scans and PDFs. You’ll build the systems that read them, pull out the right fields and tables, and hand back clean structured data that client software can trust. This role sits between OCR, layout models and LLMs, and the work is about getting the last few percent of accuracy right.

• Build extraction pipelines for scanned and digital documents, covering text, tables, checkboxes, stamps and signatures
• Combine OCR, layout-aware models and LLMs to extract fields with high accuracy
• Design validation rules and confidence scores so low-quality extractions get flagged for human review
• Build test sets from real client documents and track accuracy per field
• Work with our pharma domain experts on GMP documents, where accuracy and audit trails are mandatory

• 2+ years building document processing or OCR systems in production
• Strong Python skills
• Hands-on experience with OCR engines such as Tesseract, PaddleOCR, docTR, or cloud OCR like AWS Textract and Azure Document Intelligence
• Experience with layout-aware models such as LayoutLMv3, Donut or similar
• Experience using LLMs for structured extraction, including JSON schemas and output validation

• Experience with table extraction from complex PDFs
• Exposure to pharma, healthcare or finance documents
• Knowledge of 21 CFR Part 11 or other compliance requirements for electronic records
• Experience with vision-language models for document understanding