You'd research, build, and deploy the models that detect document fraud. Metadata tampering, visual forgeries, structural inconsistencies in PDFs. The models run in real-time and need to be fast and accurate.
Key Responsibilities:
You'd develop and train models for document verification, anomaly detection, and fraud classification. That means designing computer vision pipelines for document analysis, feature extraction, and forgery detection, then building the real-time inference services that run them at scale.
Research is part of the job. You'd read papers on document forensics and figure out what's actually usable in production (most of it isn't, but the stuff that works is gold). You'd also build training data pipelines, manage dataset curation, and set up evaluation frameworks to track model performance. You'd work with backend engineers to optimize model serving and reduce latency.
We publish when we have something worth sharing. You'd also mentor junior engineers and help decide where the team's research time goes. LLMs, vision transformers, multimodal models are all relevant to what we do.