Founded in 2002, AU10TIX is the global leader in AI driven identity verification and management, protecting the world’s largest brands against advanced fraud. The company’s future-proof product portfolio helps businesses provide frictionless customer onboarding and verification in 4-8 seconds while staying ahead of emerging threats and evolving regulatory requirements.
We are Looking for Head of Algorithms for:
- driving continuous improvement in detection rates across all algorithm domains (document fraud, deepfakes, biometrics) while ensuring production-grade performance, scalability, and reliability of deployed models
- Providing technical direction, mentorship, and career development to the Algo group
- Set the overall technical strategy and roadmap for all computer vision algorithm development
- Drive data labeling strategy and ownership across the organization, coordinate with QC, Product and various teams to define intake processes and SLAs
- Prepare and deliver technical presentations for diverse audiences: client-facing ML capability pitches, VP-level strategy decks, and internal architecture reviews
- Ensure PII compliance in algorithm pipelines and participate in cross-departmental compliance mapping initiatives
Model Development & Research
- Own and guide the design, training, and optimization of deep learning models for identity document classification, tampering detection, deepfake detection, and biometric analysis
- Lead model architecture decisions and drive migration to modern architectures
ML Lifecycle & Infrastructure
- Own the end-to-end ML pipeline: data gathering, labeling strategy, training, evaluation, versioning, and deployment
- Drive cloud migration of training pipelines to Cloud ML (compute clusters, experiment tracking, model registry, CI/CD integration)
- Oversee inference optimization: ONNX export, TensorRT FP16 acceleration, GPU benchmarking, and microservices packaging
- Define and maintain evaluation frameworks including demographic fairness testing, ROC/AUC analysis, FAR/FRR metrics, and detection rate tracking at fixed false-alarm thresholds