Senior Machine Learning Engineer, Trust

Airbnb is a global platform connecting hosts and guests through stays and experiences. The Senior Machine Learning Engineer will develop and productionize machine learning systems, AI agents, and evaluation frameworks for trust and safety, including abuse detection and autonomous trust decisions. The role owns projects end-to-end and collaborates with product, data science, engineering, fraud intelligence, and operations teams to deliver measurable impact.

Responsibilities:

• Frame and prototype ML and agentic solutions for problems that do not yet have an established approach, in partnership with product managers, data scientists, and front line defense teams
• Design, build, and productionize end-to-end Machine Learning pipelines — including feature engineering, model training, evaluation, and deployment — for both batch and real-time use cases
• Build and improve abuse behavior detection that generalizes across defenses
• Design, launch, and iterate on AI agents that automate trust decisions, including orchestration, tool interfaces, and the guardrails that hold quality steady as autonomy increases
• Build benchmarks, evaluation harnesses, and instrumentation that let us measure agentic and model decision quality objectively, and use them to drive real improvements
• Develop specialized models for trust and safety use cases, and use LLMs and AI agents to accelerate how we build models
• Write, review, and ship clean, testable code — whether training a new model, improving an existing pipeline, or optimizing a feature for scalability and reliability
• Work with large-scale structured and unstructured data to continuously improve ML models for Airbnb product, business, and operational use cases
• Partner with front line defense teams to validate solutions through experiments and holdouts, and quantify their impact on business and operational metrics
• Participate in code reviews, design discussions, and cross-team collaborations to contribute to a high-quality ML engineering culture

Requirements:

• • 5-10 years of industry experience in applied Machine Learning, with a track record of building and productionizing models at scale
• • 1-2+ years of hands-on experience with LLMs and GenAI technologies, including building with agentic frameworks, orchestration, and evaluation
• • Strong programming skills in Python (required) and familiarity with Scala, Java, or equivalent
• • Solid understanding of Machine Learning best practices — e.g., training/serving skew minimization, A/B testing, feature engineering, model selection — and algorithms such as gradient boosted trees, neural networks, transformers, and deep learning
• • Experience with ML frameworks and tooling such as TensorFlow, PyTorch, or equivalent
• • Experience with data engineering and building end-to-end ML pipelines, including both batch and real-time systems
• • Experience designing evaluation methodology for ML or LLM systems — benchmarks, ground truth, offline/online metrics, calibration
• • Comfort with ambiguity and a bias toward action: you can take a loosely defined problem, scope it, prototype quickly, and drive it to a measurable outcome
• • Exposure to architectural patterns of large, high-scale software applications (e.g., well-designed APIs, high-volume data pipelines, efficient algorithms)
• • Experience with test-driven development, incremental delivery, and deployment practices
• • A Bachelor's, Master's, or PhD in CS/ML or a related field
• This position is US - Remote Eligible. The role may include occasional work at an Airbnb office or attendance at offsites, as agreed to with your manager. While the position is Remote Eligible, you must live in a state where Airbnb, Inc. has a registered entity. Click here for the up-to-date list of excluded states. This list is continuously evolving, so please check back with us if the state you live in is on the exclusion list. If your position is employed by another Airbnb entity, your recruiter will inform you what states you are eligible to work from
• • Experience with multimodal models (vision, document, or speech) is a plus
• • Exposure to the Trust and Risk domain (e.g., fraud detection, anomaly detection, identity, account integrity) is a plus