Document Fraud Researcher

01About the role

Document forgery is evolving fast — GenAI can now produce convincing fake receipts, bank statements, and employment letters. Traditional forensic methods alone aren't enough. As a Document Fraud Researcher, you'll study emerging manipulation techniques, build adversarial test sets, and develop novel detection approaches that stay ahead of the forgery curve. Your research will directly feed into DocVerify's production models.

02What you'll do

• >Research and catalog emerging document forgery techniques — GenAI-generated docs, sophisticated splicing, metadata spoofing, and font injection.
• >Build adversarial evaluation datasets that stress-test our detection pipeline.
• >Develop novel detection signals and features for compression artifacts, rendering inconsistencies, and GenAI hallmarks.
• >Publish findings in internal reports and, where appropriate, academic venues.
• >Collaborate with the CV Engineering team to translate research into production-ready features.
• >Monitor fraud communities and dark web markets to stay ahead of emerging techniques.

03What we're looking for

• >MSc or PhD in computer science, image forensics, digital forensics, or a related field.
• >Strong research background with published work in image forensics, document analysis, or adversarial ML.
• >Hands-on experience with Python, PyTorch, and image processing libraries.
• >Deep understanding of image compression standards (JPEG, PNG, PDF rendering).
• >Ability to design controlled experiments and evaluate detection methods rigorously.
• >Strong written communication — you'll write internal research reports and external publications.

04Nice to have

• +Experience with GenAI detection (diffusion model artifacts, GAN fingerprints).
• +Background in digital forensics or law enforcement forensic analysis.
• +Familiarity with PDF internals and document rendering pipelines.
• +Experience with red-teaming or adversarial evaluation of ML systems.
• +Published at venues like CVPR, ECCV, ICDAR, or IEEE TIFS.