Spotting tampered documents
Flags edited or forged documents during customer onboarding by inspecting the image for signs of editing and cross-checking fields against each other.
- Role
- AI Systems Engineer (TechRAQ)
- Built with
- Python, FastAPI, Computer Vision, Hybrid LLM Extraction, Pydantic, Docker
Results
- Pixel/layout tamper detection
- Cross-field deterministic consistency
- Regulated KYC/onboarding
In more detail
Automated document verification and anomaly/tamper detection pipeline for regulated onboarding workflows, combining visual layout inspection with structured LLM cross-field consistency checks.
Key decisions
Visual Tamper & Font/Compression Anomaly Inspection
Analyzed document scans for Error Level Analysis (ELA) compression artifacts, bounding-box alignment shifts, and font/kerning inconsistencies indicative of digital alteration.
Structured LLM Cross-Field & Checksum Verification
Combined hybrid OCR/LLM extraction with deterministic Pydantic cross-field validators (date arithmetic, tax/ID checksums, totals reconciliation) to flag fraudulent onboarding submissions.