Senior/Staff Machine Learning Engineer (Web3)

Location: San Jose, California, United States

Salary: $223,611 – $268,333

Company

OKX is a leading crypto exchange and developer of the OKX Wallet, providing millions of users and institutions access to crypto trading and decentralized applications.

What you will do

• Design and deploy ML models for payment fraud, account takeover, scam detection, and transaction monitoring.
• Own production ML systems end-to-end, including feature pipelines, training workflows, model serving, and drift detection.
• Develop AI-powered risk capabilities such as investigation agents, case summarization, and automated decision support.
• Build production-grade LLM agent architectures using tool calling, retrieval-augmented generation (RAG), and human-in-the-loop controls.
• Partner with risk strategy, product, and operations teams to translate models into effective production controls.
• Implement evaluation frameworks for LLM agents to measure factual accuracy, latency, and operational impact.

Requirements

• Significant professional experience in ML engineering or applied data science with a track record of taking models to production.
• Strong Python skills and experience with PyTorch, TensorFlow, XGBoost, LightGBM, or scikit-learn.
• Deep knowledge of supervised learning, anomaly detection, and class-imbalanced modeling.
• Hands-on experience designing and deploying production LLM agents and agentic workflows.
• Fluency with AI-assisted engineering and the use of LLM coding tools in professional workflows.
• Must be located in or authorized to work in San Jose, California, United States.

Nice to have

• Experience building AI agents for fraud, risk, compliance, customer operations, or cybersecurity domains.

Culture & Benefits

• Competitive base salary, performance bonuses, and long-term incentives.
• Comprehensive medical and financial benefits package.
• Inclusive work environment driven by principles of "We Before Me" and "Get Things Done".
• Opportunity to work with unique, high-scale datasets spanning on-chain and off-chain activity.