OCR / Machine Learning Engineer (Document AI) - Fileme

About the role
FIleme is building an enterprise-grade document extraction engine for receipts, invoices, remittances, bank statements, and financial documents. We’re developing a multi-stage OCR/ML pipeline inspired by platforms like Dext Prepare — with a focus on accuracy, automation, and seamless integration with accounting ecosystems.

We’re looking for an OCR / Machine Learning Engineer to join on a contract or part-time basis to help build and scale our next-generation extraction engine. You will design and implement models, pipelines, and cloud services that turn messy, real-world financial documents into structured financial data.

This role is ideal for someone who thrives on technical autonomy, enjoys solving complex document-AI problems, and has prior experience with financial OCR, document intelligence, or applied deep learning.

Responsibilities

• Build extraction pipelines for receipts, invoices, remittance advices, ATM slips, bank statements, and supporting financial documents.
• Implement preprocessing (deskewing, denoising, segmentation, contour detection, multi-language text normalization).
• Integrate and evaluate OCR engines (open-source + cloud APIs such as Tesseract, PaddleOCR, Datalabs/Marker, AWS Textract).
• Train or fine-tune ML/DL models (e.g., classification, key-value extraction, layout analysis, entity detection).
• Evaluate accuracy across document types and propose improvements.
• Build cloud-based inference endpoints (AWS preferred).
• Work with the engineering team to integrate extraction output into the ExpenseHub staging pipeline.
• Create validation and error-handling logic for ambiguous outputs.
• Implement analytics and reporting on extraction accuracy, throughput, and edge-case performance.

Qualifications

• 5+ years of experience in OCR, Computer Vision, Machine Learning, or Document AI.
• Strong Python engineering experience including async services, model lifecycle, and API integration.
• Hands-on experience with deep learning frameworks (PyTorch preferred).
• Experience with structured extraction of financial documents is a strong advantage.
• Understanding of document layouts, table detection, and field extraction.
• Experience deploying ML systems to AWS (Lambda, S3, EC2, ECS, API Gateway).
• Strong understanding of accuracy metrics, error profiles, and evaluation frameworks for OCR.
• Ability to work independently and deliver high-quality, production-ready code.

Nice to have

• Experience with Xero, QuickBooks, or accounting/fintech platforms.
• Experience with multi-tenant SaaS architectures.
• Experience with JSON schema versioning, validation frameworks, and ETL pipelines.
• Familiarity with Datalabs, Dext Prepare, Rossum, Veryfi, or Glean.