About the Role
You will own the full IDV ML stack including document authenticity models, 1:1 and 1:N face-match, liveness detection, presentation-attack detection, and deepfake/injection detection from feature pipeline through threshold tuning and production enforcement. You will build identity graph systems using GNNs that cluster accounts sharing biometric, device, and document signals to detect synthetic-identity rings and coordinated fraud at onboarding. You will develop behavioral and device intelligence models for capture-session anomaly detection, bot-vs-human classification, and device-fingerprint-based risk scoring at real-time latency. You will drive vendor ML strategy by benchmarking external models against an evaluation set owned by Coinbase, designing dynamic routing logic across providers and geographies, and building the in-house evaluation layer that catches regressions before they reach users. You will lead and mentor senior and mid-level engineers in the pod while partnering with ML Platform and Risk ML teams to align cross-company ML system design.
Requirements
- 8+ years deploying production ML systems at scale, with proven technical leadership owning cross-team ML architecture from design through production
- Domain experience in identity verification, biometrics, or account integrity with deep applied ML in at least two of computer vision, biometrics, GNNs, sequence models, or NLP/LLMs
- Expert-level Python with production experience in TensorFlow or PyTorch, including model training, evaluation, and serving infrastructure
- Track record translating KYC/AML requirements and fraud trends into ML roadmaps and communicating trade-offs to Product, Compliance, Risk, and Security stakeholders
- Utilizes generative AI responsibly, maintaining human oversight to deliver business-ready outputs and drive measurable improvements in workflow efficiency, cost, and quality
Responsibilities
- Own the full IDV ML stack including document authenticity models, 1:1 and 1:N face-match, liveness detection, presentation-attack detection, and deepfake/injection detection from feature pipeline through threshold tuning and production enforcement
- Build identity graph systems using GNNs that cluster accounts sharing biometric, device, and document signals to detect synthetic-identity rings and coordinated fraud at onboarding
- Develop behavioral and device intelligence models for capture-session anomaly detection, bot-vs-human classification, and device-fingerprint-based risk scoring at real-time latency
- Drive vendor ML strategy by benchmarking external models against an evaluation set owned by Coinbase, designing dynamic routing logic across providers and geographies, and building the in-house evaluation layer that catches regressions before they reach users
- Lead and mentor senior and mid-level engineers in the pod while partnering with ML Platform and Risk ML teams to align cross-company ML system design