Senior/Staff Machine Learning Engineer

OKX•San Jose, CA

9h•$223,611 - $268,333

About The Position

Building machine learning systems for risk at a global crypto exchange is fundamentally different from conventional ML engineering. The data spans on-chain activity, fiat deposits and withdrawals, trading behaviour, account access patterns, device intelligence, identity information, and customer interactions—signals that very few organizations can analyze together. The problems are complex, adversarial, and constantly evolving. Models must identify emerging fraud patterns, scams, account takeovers, payment abuse, and other forms of financial risk while minimizing disruption to legitimate customers. Success is not measured only through offline model metrics. It is measured through prevented losses, improved approval rates, reduced false positives, faster investigations, and more reliable customer experiences. This role sits within a multidisciplinary risk team of machine learning engineers, data scientists, risk strategy specialists, analytics engineers, product managers, and operations teams. You will work across the full ML lifecycle—from problem formulation, feature engineering, and model development to real-time deployment, monitoring, experimentation, and continuous iteration. You will also help shape how AI is used across the risk organization. LLM-assisted development, automated model workflows, AI-powered investigations, and intelligent review agents are part of the team’s daily work. We are looking for engineers who already use these tools effectively and can help establish safe, scalable, and production-ready AI practices.

Requirements

Significant professional experience in machine learning engineering, applied data science, or a closely related field, with a strong record of taking models from prototype to production. Scope and level will be calibrated based on experience.
Strong Python skills and hands-on experience with machine learning frameworks such as PyTorch, TensorFlow, XGBoost, LightGBM, or scikit-learn.
Strong knowledge of applied machine learning fundamentals, including supervised learning, anomaly detection, representation learning, class-imbalanced modeling, model calibration, and evaluation under changing data distributions.
Demonstrable fluency with AI-assisted engineering. You regularly use LLM coding tools, have built AI-integrated workflows or applications, and understand both the productivity benefits and the security, reliability, and governance risks.
Familiarity with model explainability techniques such as SHAP, feature attribution, reason-code generation, and model scorecards.
Hands-on experience designing and deploying production LLM agents, including agentic workflows, tool calling, retrieval-augmented generation, prompt and context management, structured output generation, and multi-step task orchestration.
Experience integrating LLM agents with internal systems, APIs, databases, search tools, case-management platforms, or decision engines to automate complex operational workflows.
A strong understanding of LLM-agent evaluation and reliability, including hallucination control, grounding, observability, permissions, failure handling, human review, latency, and cost optimization.
Strong communication and collaboration skills, with the ability to work effectively with engineers, data scientists, risk specialists, product managers, operations teams, and legal or compliance stakeholders.

Nice To Haves

Experience building AI agents for fraud, risk, compliance, customer operations, cybersecurity, or other high-stakes domains is a meaningful advantage.

Responsibilities

Design, build, and deploy machine learning models for risk use cases such as payment fraud, account takeover, scam detection, deposit and withdrawal risk, promotional abuse, customer risk assessment, and transaction monitoring.
Own production ML systems end to end, including feature pipelines, training workflows, model serving, decision integrations, monitoring, alerting, drift detection, retraining, and incident response.
Partner with risk strategy and product teams to translate models into effective production controls, including approval, rejection, review, cooldown, limit adjustment, account restriction, and other risk mitigation actions.
Work closely with risk operations teams to understand investigation workflows, incorporate reviewer feedback, improve model explainability, and continuously refine labels and training data.
Apply AI-assisted development throughout the engineering workflow, using LLM coding tools to accelerate implementation, testing, debugging, analysis, and documentation while maintaining appropriate security and review standards.
Develop AI-powered risk capabilities such as investigation agents, case summarization, evidence collection, review recommendations, alert triage, suspicious-entity mining, and automated decision support.
Take research-stage models into reliable production systems by validating feature logic, reviewing data quality, addressing latency and scalability constraints, and ensuring consistency between offline training and online inference.
Ensure models and decision systems are explainable, traceable, and well documented so that model outputs can be understood by risk operations, product stakeholders, internal governance teams, and regulators where applicable.
Design, build, and deploy LLM-based agents for risk operations and investigation workflows, including case triage, evidence retrieval, transaction analysis, alert summarization, review recommendations, and automated action orchestration.
Develop production-grade agent architectures using tool calling, retrieval-augmented generation, workflow orchestration, structured outputs, memory, guardrails, and human-in-the-loop controls.
Build evaluation frameworks for LLM agents, measuring factual accuracy, task completion, decision consistency, latency, cost, reviewer acceptance, and operational impact. Ensure LLM agents operate safely in a regulated risk environment by implementing permission controls, audit logs, data privacy protections, prompt and tool security, fallback mechanisms, and clear escalation paths.